Fascin1 suppresses RIG-I–like receptor signaling and interferon-β production by associating with IκB kinase ϵ (IKKϵ) in colon cancer

Abstract

Fascin1 is an actin-bundling protein involved in cancer cell migration and has recently been shown also to have roles in virus-mediated immune cell responses. Because viral infection has been shown to activate immune cells and to induce interferon-β expression in human cancer cells, we evaluated the effects of fascin1 on virus-dependent signaling via the membrane- and actin-associated protein RIG-I (retinoic acid-inducible gene I) in colon cancer cells. We knocked down fascin1 expression with shRNA retrovirally transduced into a DLD-1 colon cancer and L929 fibroblast-like cell lines and used luciferase reporter assays and co-immunoprecipitation to identify fascin1 targets. We found that intracellular poly(I·C) transfection to mimic viral infection enhances the RIG-I/MDA5 (melanoma differentiation-associated gene 5)-mediated dimerization of interferon regulatory factor 3 (IRF-3). The transfection also significantly increased the expression levels of IRF-7, interferon-β, and interferon-inducible cytokine IP-10 in fascin1-deleted cells compared with controls while significantly suppressing cell growth, migration, and invasion. We also found that fascin1 constitutively interacts with IκB kinase ϵ (IKKϵ) in the RIG-I signaling pathway. In summary, we have identified fascin1 as a suppressor of the RIG-I signaling pathway associating with IκB kinase ϵ in DLD-1 colon cancer cells to suppress immune responses to viral infection.

Introduction

Fascin1 is an evolutionarily conserved 55-kDa actin-bundling protein that is localized in filopodia, which are involved in cell migration, as well as actin-based protrusions beneath the plasma membrane (1, 2). Fascin1 protein plays an important role in the regulation of cell adhesion, migration, and invasion in breast cancer (3), colon cancer (4), and melanoma (5) and is closely correlated with a poor prognosis and survival in cancer patients. Because the inhibition of fascin1 function leads to a block in tumor metastasis (6, 7), it is considered a potential therapeutic target for tumors. On the other hand, fascin1 is also reportedly involved in immune responses, such as cytokine production (8).

The discovery of tumor cells mimicking an innate immunity pathway after viral infection has attracted considerable attention. For example, the activation of RIG-I (retinoic acid-inducible gene 1) and MDA5 (melanoma differentiation-associated gene 5), both of which are viral RNA recognition molecules, in tumor cells can induce immunogenic cell death in melanoma (9) and leukemia cells (10). A synthetic analog of dsRNA, poly(I·C), activated the pathway mediated by MDA5/RIG-I/IFN-regulatory factor (IRF)-3³ to express immunoadjuvant IFN-β in PC3 and DU145 cell lines derived from a prostate cancer metastasis (11) and in pancreatic cancer cells (12). Regarding colon cancer, dsRNA induced by demethylation activated MDA5 and elicited the downstream activation of interferon promoter stimulator-1 (IPS-1), IRF-7, and type III IFNs and the up-regulation of IFN-responsive genes in a LIM1215 colorectal cancer cell line (13).

RIG-I, which is reportedly localized in membrane ruffles, associates with the actin cytoskeleton in the human colon adenocarcinoma cell line HT-29 (14) and in the immortalized human intestinal cell line Caco-2 (15). The reorganization of the actin cytoskeleton leads to RIG-I relocalization and induces type I IFNs in Caco-2 cells (15). These findings have culminated in the hypothesis that fascin1 is involved in the RIG-I pathway in colon cancer.

We herein investigated the effects of fascin1 on cell growth and invasion in the RIG-I immune pathway by stimulating the DLD-1 colon cancer cell line with poly(I·C). To our knowledge, this is the first report to clarify the role of fascin1 in dsRNA-dependent RIG-I signaling in the promotion of metastasis in cancer cells, a function that is distinct from the protein's known cytoskeletal properties.

Results

Knockdown of fascin1 suppressed cell migration and invasion

To examine the roles of fascin1 in tumor cells, we first stably deleted fascin1 using shRNA in DLD-1 cells and L929 fibroblast-like cells. The expression levels of mRNA and protein in fascin1 shRNA-transfected cells were effectively decreased to 25 and 20% of those in the control cells, respectively, in DLD-1 cells (Fig. 1A). Consistent with a previous report (16), a morphological analysis elucidated that filopodia formation, but not lamellipodia formation, was impaired in fascin1-deleted L929 cells (Fig. 1B), indicating that fascin1 is a specific actin cross-linker that provides stiffness for filopodial bundles. Regarding cell growth, no difference was observed between DLD-1 cells with or without fascin1 (Fig. 1C), whereas the knockdown of fascin1 suppressed DLD-1 cell migration and invasion (Fig. 1D). These findings showed that fascin1 promoted cell motility in DLD-1 cells. Fascin1-mediated filopodia assembly has been reported to be dispensable for mouse development, and fascin1-deficient mice are viable and fertile with no major developmental defects (17).

Figure 1.

Knockdown of fascin1 suppressed migration and invasion but not cell growth. A, stable knockdown of fascin1 in DLD-1 and L929 cells was confirmed using real-time RT-PCR (n = 2) and an immunoblot (IB) analysis. B, F-actin was stained with rhodamine-phalloidin, and images of filopodia in L929 cells were obtained using confocal microscopy. Original magnification, ×2560. C, cell growth was measured using the MTS assay 48 h after the inoculation of DLD-1 cells (n = 3). D, cell migration (n = 3) and invasion (n = 3) of DLD-1 cells were examined using the Boyden-chamber assay. *, p < 0.05; ***, p < 0.001.

Knockdown of fascin1 enhanced RIG-I/MDA5-mediated IRF-7 and IFN-β induction by poly(I·C) transfection

We focused on two receptors for viral RNA, TLR3 (Toll-like receptor 3) and RIG-I–like receptors. IFN-β mRNA production was similar in DLD-1 cells with or without fascin1 when poly(I·C) was applied extracellularly (Fig. 2A), implying no involvement of fascin1 in the TLR3 pathway. The expression of TLR3 in DLD-1 cells has been confirmed by Sadik et al. (18), and we also have data showing the expression of mRNA (data not shown). Although the dose of poly(I·C) was increased from 5 μg/ml (equivalent to the intracellular dose) to 30 μg/ml, extracellularly administered poly(I·C) did not cause IFN-β mRNA expression to differ between control cells and fascin1-deleted cells. This result was supported by another poly(I·C) performed at a dose of 5 μg/ml to 100 μg/ml (data not shown).

Figure 2.

Knockdown of fascin1 enhanced RIG-I/MDA5-mediated gene induction by the poly(I·C) transfection of DLD-1 and L929 cells. A, poly(I·C) was applied extracellularly for 6 h, and the mRNA expression of IFN-β was measured using real-time RT-PCR. B and C, poly(I·C) was transfected intracellularly for 6 h (B), and the mRNA expressions of IFN-β and IP-10, as well as those for RIG-I, MDA5, IRF-7, and IFIT1s, were measured using real-time RT-PCR (C, n = 6). D, the re-expression of fascin1 using p-FLAG-CMV4 vector in fascin1-deleted cells was confirmed using real-time RT-PCR (n = 2). The alteration of IFN-β mRNA expression by knockdown and the re-expression of fascin1 under poly(I·C) were examined using real-time RT-PCR (n = 3). *, p < 0.05; **, p < 0.01.

When poly(I·C) was transfected intracellularly, however, the expression levels of IFN-β and IP-10 (interferon-γ–inducible protein-10) mRNA were significantly increased in fascin1-deleted DLD-1 cells, with no difference observed without poly(I·C) treatment. These results were supported by L929 cell assays (Fig. 2B). Furthermore, the mRNA expression levels of RIG-I–, MDA5–, IRF-7–, and IFN-stimulated genes including IFIT (IFN-induced protein with tetratricopeptide repeats) 1, 2, and 3 were increased by poly(I·C) transfection in fascin1-depleted DLD-1 cells (Fig. 2C), indicating that fascin1 suppressed IRF-7 and IFN-β expression levels via the RIG-I/MDA5-dependent signaling pathway by poly(I·C) stimulation. To confirm the suppressive effect of fascin1 on IFN-β mRNA expression, we conducted a rescue experiment by the re-expression of fascin1 in fascin1-deleted DLD-1 cells treated with poly(I·C). The increase in IFN-β mRNA expression arising from fascin1 knockdown was significantly suppressed by the re-expression of fascin1 (Fig. 2D).

Knockdown of fascin1 enhanced IRF-3 activation

Because transcription of the IFN-β gene requires the activation of IRF-3, NF-κB, and several other transcription factors (19), we assessed whether fascin1 contributed to the activation of IRF-3 or NF-κB/p65. As shown in Fig. 3A, a poly(I·C) time-dependent definite increase in IRF-3 phosphorylation was detected in fascin1-deleted DLD-1 cells, compared with control cells, using an immunoblot analysis. The dimerization of IRF-3 at 6 h after poly(I·C) transfection supported this result (Fig. 3B). The phosphorylation of the NF-κB subunit p65/RelA was examined next. The amount of phosphorylated p65/RelA protein increased in a time-dependent manner in response to poly(I·C) transfection; however, the levels were comparable between fascin1-deleted cells and controls (Fig. 3C). These data suggest that fascin1 suppresses the activation of IRF-3 but not the activation of the NF-κB signaling pathway.

Figure 3.

Knockdown of fascin1 enhanced poly(I·C)-induced IRF-3 activation in DLD-1 cells. A, the phosphorylation of IRF-3 was examined using an immunoblot (IB) analysis. B, IRF-3 dimerization was detected using native-PAGE. The expressions of IRF-3 and fascin1 in whole cell lysates were examined using an immunoblot analysis. C, the phosphorylation of the NF-κB subunits p65/RelA in whole cell extracts was analyzed using an immunoblot analysis. The results are representative of three independent experiments. WCL, whole cell lysates.

Fascin1 constitutively associated with IKKϵ in RIG-I signaling

The knockdown of fascin1 was found to enhance IRF-3 activation and the production of IRF-7, IFN-β, and IFITs via the RIG-I/MDA5 signaling pathway (Fig. 2, B and C). Therefore, we conducted IFN-β luciferase reporter assays to further investigate the molecules affected by fascin1. Fascin1 dose-dependently inhibited IFN-β luciferase reporter activity induced by the overexpression of RIG-I, IPS-1, TBK1 (TANK-binding kinase-1), and IKKϵ, but not that by IRF-3(5D) (Fig. 4A). The inhibition of IKKϵ–induced IFN-β by fascin1 overexpression was slight but dose-dependent, indicating an association between fascin1 and IKKϵ. Together with IRF-3, which did not alter IFN-β production, fascin1 was considered to abrogate IFN-β production by interfering with the function of IKKϵ. To confirm that fascin1 interacted with IKKϵ, we next performed co-immunoprecipitation experiments with combinations of vector plasmids expressing Myc-tagged fascin1 (pcDNA3) and FLAG-tagged TBK1, IKKϵ, or IRF-3 (pFLAG-CMV4). As expected, IKKϵ and IRF-3, but not TBK1, co-immunoprecipitated with fascin1 (Fig. 4B). To verify these results, we conducted reciprocal co-immunoprecipitation assays. As shown in Fig. 4C, fascin1 co-precipitated with IKKϵ, but not with TBK1 or IRF-3. Thus, fascin1 was confirmed to constitutively interact with IKKϵ, rather than with RIG-I, IPS-1, TBK1, or IRF-3.

Figure 4.

Targets associated with fascin1 were analyzed using HEK293T cells. A, targets of fascin1 were evaluated using IFN-β luciferase reporter assays (n = 3). B and C, targets of fascin1 were confirmed using a co-immunoprecipitation (IP) analysis (B) and a reciprocal co-immunoprecipitation analysis (C). The data are representative of three independent experiments. D, construction of full-length IKKϵ (complete) and its deletion mutants. The essential domain of IKKϵ required for its association with fascin1 was detected using a co-immunoprecipitation analysis. The data are representative of three independent experiments. E, domains of fascin1 required for its association with IKKϵ were examined using a co-immunoprecipitation analysis. The data are representative of two independent experiments. F, IFN-β suppressive activity of three fascin1 mutants were evaluated using luciferase reporter assays (n = 3). G, co-immunoprecipitation of endogenous fascin1 and IKKϵ in DLD-1 cells with or without poly(I·C). H, influence of fascin1 knockout on mRNA level of TBK1 and IKKϵ in DLD-1 cells (n = 3). *, p < 0.05; **, p < 0.01; ***, p < 0.001. IB, immunoblot; WCL, whole cell lysates.

Human IKKϵ contains a kinase domain, an ubiquitin-like domain, a leucine zipper, and a helix–loop–helix region (20). To better characterize fascin1/IKKϵ interactions, deletion mutants of these domains were engineered, and co-immunoprecipitation analyses were performed. Fascin1 only associated with deletion mutants containing the kinase domain region (Fig. 4D). Fascin1 is also well-known to bind F-actin and PKC-α. To evaluate whether F-actin or PKC-α–binding sites of fascin1 were essential for IKKϵ associations, we constructed fascin1 S39A (PKC-α–binding site), fascin1 H392A, and fascin1 A488W (F-actin–binding site) mutants for co-immunoprecipitation assays with IKKϵ. IKKϵ associated with all of these mutants (Fig. 4E), indicating that the kinase domain of IKKϵ was required for binding to fascin1 in a manner independent of the protein's F-actin– or PKC-α–binding domains. These three fascin1 mutants inhibited IFN-β luciferase reporter activity induced by the overexpression of RIG-I, similar to the actions of WT fascin1 shown in Fig. 4A, indicating that the mutation of fascin1 did not influence IFN-β production (Fig. 4F). Poly(I·C) strengthened the association of endogenous fascin1 and IKKϵ in DLD-1 cells, as shown in Fig. 4G. This result suggests that fascin1 might act as an initial brake on RLR signaling that is released by stimulation but not as a negative feedback regulator, enabling the interaction to increase after stimulation. Binding site characterization was performed using HEK293T cells instead of DLD-1 cells, because the co-immunoprecipitation of fascin1/IKKϵ with overexpressed IKKϵ was less pronounced in the DLD-1 cell line. Because the mRNA levels of TBK1 and IKKϵ without poly(I·C) treatment were similar between the control and fascin1-deleted DLD-1 cells, fascin1 deletion was confirmed to have no effect on the level of RIG-I–signaling molecules (Fig. 4H).

Poly(I·C) suppressed growth, migration, and invasion in fascin1-deleted DLD-1 cells

Next, the anti-cancer effect of fascin1 knockdown on RIG-I signaling was investigated. The knockdown of fascin1 did not alter cell growth (Fig. 5A), as shown in Fig. 1C, but growth in fascin1-deleted cells was markedly suppressed by poly(I·C) transfection, compared with PBS-only treatment. This finding was consistent with the augmented cell death (16–22%) observed by poly(I·C) transfection in fascin1-deleted cells (Fig. 5B). Cell migration and invasion were inhibited by fascin1 knockdown (Fig. 5C), similar to the findings shown in Fig. 1D, and were more strongly suppressed by poly(I·C) transfection, compared with PBS treatment, in fascin1-deleted cells (Fig. 5C). We clarified that the impaired migration ability induced in fascin1-deleted DLD-1 cells under poly(I·C) treatment depended on the RIG-I mediated signaling pathway by the knockdown of RIG-I using siRNA (Fig. 5D). These results suggest that fascin1 binding to IKKϵ prevents cancer cell growth, migration, and invasion via RIG-I signaling.

Figure 5.

Transfection of poly(I·C) into fascin1-deleted DLD-1 cells suppressed cell growth, migration, and invasion. A and B, 48 h after poly(I·C) transfection, cell growth was measured using the MTS assay (A, n = 6), and apoptosis was determined using FACS 24 h later (B). C, cell migration (n = 3) and invasion (n = 3) were examined using the Boyden-chamber assay. D, knockdown of RIG-I was confirmed using real-time RT-PCR (n = 2), and cell migration was compared between si-control and si-RIG-I–treated fascin1-deleted cells with or without poly(I·C) transfection (n = 3). *, p < 0.05; **, p < 0.01; ***, p < 0.001. 7-AAD, 7-aminoactinomycin D.

Knockdown of fascin1 enhanced the phosphorylation of Ser-727 in STAT1 by IFN-β treatment

To clarify the effects of fascin1 on the IFN feedback loop, the phosphorylation of signal transducer and transcription activator 1 (STAT1) was evaluated in DLD-1 cells. In response to IFN-β treatment, the phosphorylation of Tyr-701 in STAT1 was activated in fascin1-deleted cells using the same kinetics as those occurring in control cells. In contrast, the phosphorylation of Ser-727 in STAT1 was remarkably enhanced in fascin1-deleted cells (Fig. 6A). Although the phosphorylation of Ser-727 in STAT1 has been shown to require p38 MAPK activity (21) and to be regulated by CDK8 (22), neither of these were related to fascin1 (data not shown). These findings implied that fascin1 modulated the feedback loop of the IFN-β signaling pathway by suppressing STAT1 Ser-727 phosphorylation. Based on the above findings, the role of fascin1 in the virus-dependent RIG-I signaling pathway (Fig. 6B, left panel) and its IFN feedback loop (Fig. 6B, right panel) in DLD-1 colon cancer cells was schematically proposed in Fig. 6B.

Figure 6.

Knockdown of fascin1 enhanced STAT1 Ser-727 phosphorylation in the IFN-β feedback loop in DLD-1 cells. A, The phosphorylation of Tyr-701 and Ser-727 in STAT1 was analyzed using an immunoblot (IB) analysis after stimulation with 1 ng/ml of IFN-β for the indicated times. The data are representative of two independent experiments. B, proposed schematic of fascin1 function in poly(I·C)-dependent RIG-I signaling (left panel) and the IFN feedback loop (right panel).

Discussion

In addition to activating dendritic cells, viral infection has also been found to induce endogenous IFN-β in human cancer cells (23). Fascin1 is involved in immune responses such as cytokine production (24). The present study clearly demonstrated the involvement of fascin1 in the dsRNA-dependent RIG-I signaling pathway of DLD-1 colon cancer cells upon viral stimulation. Whereas TLRs are located on the cell surface or in endosomes, where they recognize viral RNA and DNA, RIG-I and MDA5 are cytosolic nucleic acid sensors that recognize viruses intracellularly (24). Here, DLD-1 cells with or without fascin1 did not display differential responses to extracellular poly(I·C) treatment as evaluated using the IFN-β mRNA level. However, the mRNA levels of RIG-I, MDA5, and IFN-β were significantly increased by intracellular poly(I·C) transfection in fascin1-deleted cells, indicating that the knockdown of fascin1 enhanced IFN-β production via a RIG-I-dependent pathway and not via TLR3-mediated responses in DLD-1 cells.

RIG-I and MDA5 interact with IPS-1 leading to the recruitment of TRAF3 (TNF receptor–associating factor 3), TBK1, and IKKϵ, causing the nuclear translocation of IRF-3 or leading to the activation of IKKγ, IKKα, and IKKβ, resulting in the activation of NF-κβ (19, 24). We examined the pathways that fascin1 recognizes as its target (i.e. the IKKϵ/IRF-3 pathway and the IKKγ/IKKα/NF-κβ pathway) under poly(I·C) transfection. No alterations were observed in the activation of NF-κβ/p65 between DLD-1 cells with or without fascin1, although dimerization of IRF-3 was evident in fascin1-deleted cells. IRF-3 dimerization is associated with the activation and translocation of IRF-3 from the cytoplasm to the nucleus (25) to induce IFN-β. This confirmed the influence of fascin1 on the IKKϵ/IRF-3 pathway by poly(I·C) transfection in DLD-1 cells.

We next searched for fascin1 associations with RIG-I, IPS-1, TBK1, IKKϵ, and IRF-3. We concluded that fascin1 interacted with IKKϵ rather than with RIG-I, IPS-1, TBK1, or IRF-3, using a luciferase reporter assay and co-immunoprecipitation; however, the mechanism by which IKKϵ–induced IFN-β induction is impaired by fascin1 is completely unknown. One study revealed that the activation of IKKϵ is rapid and transient and precedes a more prolonged activation of TBK1 in immune cells (26). Even though IKKϵ and TBK1 are generally treated as a complex, TBK1 is reported to be predominantly responsible for IRF3 activation, and the involvement of IKKϵ in the relation between TBK1 and IRF-3 remains uncertain (27). Capturing the behavior of IKKϵ may be difficult. Thus, the mechanism by which fascin1 impairs the induction of IFN-β by associating with IKKϵ is likely to be complex but needs to be resolved. IKKϵ influences cell proliferation and transformation and is thereby classified as an oncogene (28); meanwhile, the overexpression of IKKϵ has been documented to activate IRF-3 (29), which induces IFN-β production by viral infection. In our experiments, IKKϵ without an association with fascin1 was considered to promote the production of IRF-7, IFN-β, IFITs, and IP-10 to suppress cell growth and invasion under poly(I·C) treatment. The diminished cell growth was accompanied by an increase in apoptosis. Regarding its inhibitory effect on growth, poly(I·C) has been reported to cause apoptosis resulting from ROS-mediated DNA damage and IFN-β-mediated growth arrest in renal cell carcinoma (30). However, the mechanism by which the knockdown of fascin1 enhances apoptosis remains unclear.

As for the inhibitory effect of fascin1 knockdown on cell invasion, we observed a significant increase in IRF-7 expression in fascin1-deleted cells. According to Li et al. (31), highly expressed miR-762 in breast cancer cell lines promoted cell invasion by down-regulating IRF-7. Furthermore, patients exhibiting a high expression of IRF-7–regulated genes in primary tumors were associated with prolonged bone metastasis-free survival, and the restoration of IRF-7 or the administration of IFN to breast cancer cells led to a reduction in bone metastases (32). Thus, the inhibitory effect of fascin1 knockdown on invasion in DLD-1 cells might be regarded as depending largely on an increase in IRF-7. Concerning INF-β and IP-10, the former inhibited the proliferation and invasion of pancreatic cancer cells (33), and the latter was involved in the induction of apoptosis and the inhibition of colorectal cancer growth (34). On the other hand, excreted endogenous type I IFN recruits and activates CD8α+ dendritic cell populations, resulting in tumor regression (35) as shown in Fig. 6B. Therefore, the inhibition of IFN production is preferable for tumor cell survival, which may be, in part, mediated by fascin1.

Because IKKϵ directly phosphorylates STAT1 (36) and exerts antitumor activity (37, 38), we investigated how fascin1 affects STAT1 phosphorylation in the IFN feedback loop. RIG-I facilitates the positive feedback loop of type I IFN production (39). In this pathway, IFN-β stimulates the JAK-STAT pathway leading to the phosphorylation of STAT1 and STAT2 (40, 41). STAT1/2 heterodimers and IRF9 form a protein complex that participates in this feedback loop (42). In our experiments with DLD-1 cells, IFN-β increased the phosphorylation of Tyr-701 in STAT1, similar to the results of another report (43), although fascin1 knockdown did not affect the degree of phosphorylation. Perwitasari et al. (44) reported that IKKϵ phosphorylated Ser-708 in STAT1 by IFN-β treatment, whereas Zhang et al. (45) dismissed this possibility in T cells, suggesting that IKKϵ relayed signaling depending on the cell type and apical signal. This phosphorylation site of IKKϵ in IFN-β signaling is believed to exhibit diversity. In the current study, the phosphorylation of Ser-727 in STAT1 was enhanced in fascin1-deleted cells, indicating the possibility of the suppressive phosphorylation control of Ser-727 in STAT1 by fascin1. The association of fascin1 with IKKϵ is most likely involved in the phosphorylation of Ser-727 in the IFN-β feedback loop in DLD cells. Taken together, our results provide evidence that fascin1 not only participates in the initial IFN-β production stage but is also involved in the IFN signaling feedback loop by associating with IKKϵ, as proposed in Fig. 6B. Identifying the roles of the fascin1/IKKϵ association on STAT1 in the feedback loop will be a major issue in future research. In conclusion, we have identified fascin1 as a suppressor of the RIG-I signaling pathway associated with IKKϵ in DLD-1 colon cancer cells, resulting in the suppression of immune responses to viral infection. Further study of the possible therapeutic application of fascin1 is warranted.

Experimental procedures

Cell culture

DLD-1, L929 and HEK293T cells were maintained in Dulbecco's modified Eagle's medium (Sigma–Aldrich). All the cultures were supplemented with 10% heat-inactivated fetal bovine serum and penicillin/streptomycin.

Antibodies and reagents

Antibodies against phospho-IRF-3 (Ser396), NF-κB/p65, phospho-NF-κB/p65 (S536), p38 MAPK, phosphor-p38 MAPK (Thr180/Tyr182), STAT1, phosphor-STAT-1 (Tyr701) and phosho-STAT1 (Ser727) were purchased from Cell Signaling Technology (Danvers, MA); antibodies against β-actin, fascin1 and IRF-3 were obtained from Santa Cruz Biotechnology (Dallas, TX); antibodies against DDDDK-tag and Myc-tag were procured from MBL (Nagoya, Japan); and secondary antibodies were obtained from Invitrogen (Carlsbad, CA). Annexin V Alexa 647 (BioLegend, San Diego, CA), 7-aminoactinomycin D (Enzo Life Science, Farmingdale, NY), rhodamine phalloidin (Cytoskeleton, Inc., Denver, CO), poly(I·C) (Sigma–Aldrich and InvivoGen, San Diego, CA; used at a 5 μg/ml concentration), pSINsi-hU6 vector (TaKaRa Bio, Inc., Shiga, Japan), pRL-TK vector (Promega, Madison, WI), and pcDNA3 and pFLAG-CMV4 vectors (Invitrogen) were also used.

Retroviral transduction

The fascin1 knockdown vector was constructed using the pSINsi-hU6 shRNA expression retrovirus vector. We used two target sequences of human fascin1, as follows: 5′-GGTGGGCAAGGACGAGCTC-3′ and 5′-GCCTGAAGAAGAAGCAGAT-3′ (6). DLD-1 or L929 cells stably expressing fascin1-shRNA were selected in 500 μg/ml of G418-containing medium.

Knockdown of RIG-I using siRNA

The target sequences for the two pairs of human RIG-I gene were as follows: control, sense 5′-UUCUCCGAACGUGUCAACGUTT-3′ and antisense 5′-ACGUGACACGUUCGGAGAATT-3′; RIG-I #1, sense 5′-UAAGGUUGUUCACAAGAAUCUGUGG-3′ and antisense 5′-CCACAGAUUCUUGUGAACAACCUUA-3′; and RIG-I #2, sense 5′-UGGACAUGAAUUCUCACUAAGAUUC-3′ and antisense 5′-GAAUCUUAGUGAGAAUUCAUGUCCA-3′. DLD-1 cells were transfected with siRNAs (10 nm) using Lipofectamine RNAi MAX Reagent (Thermo Fisher Scientific Inc.).

Immunofluorescence

Cells were fixed with 4% PFA in PBS and permeabilized with Triton X-100. After blocking, the cells were incubated with primary antibodies overnight at 4 °C, followed by incubation with secondary antibodies for 60 min. A confocal microscopy analysis was performed using a Leica TCS SP2 AOBS (Leica Microsystems, Wetzlar, Hessen, Germany).

Boyden-chamber cell migration assay

A total of 1 × 10⁵ cells suspended in 500 μl of serum-free medium were plated onto the cell culture inserts (8-μm pore size; Becton Dickinson, Franklin Lakes, NJ), which were then placed into 24-well dishes containing 10% fetal bovine serum culture medium. The cells in the upper chamber were removed using a cotton swab 20–24 h later, and the migrated cells were fixed in 100% methanol and stained with crystal violet solution (0.2% crystal violet in 2% ethanol) for 10 min. Three randomly selected fields on the lower side of the inserts were photographed, and the migrated cells were counted.

Cell invasion assay

The cell invasion assays were identical to the cell migration assay apart from the precoating of the cell culture insert with 2 mg/ml of Matrigel^TM matrix basement membrane (Becton Dickinson) for 4–5 h at 37 °C. The invasion period was 40–48 h.

Real-time RT-PCR

Total RNA was prepared using RNAiso plus (TaKaRa Bio, Inc.), and cDNA was synthesized using PrimeScript RT Master Mix (TaKaRa Bio, Inc.). A Thermal Cycler Dice^TM real-time system (TaKaRa Bio, Inc.) was used. All the data were normalized to the expression of GAPDH mRNA and were presented as the fold-increase relative to the control cells. The primer sets that were used are listed in Table 1.

Table 1.

Primers used for real-time RT-PCR

Gene		Sequence
hu GAPDH	Forward	5′-CACCACCAACTGCTTAGCACC-3′
	Reverse	5′-CAGTCTTCTGGGTGGCAGTGATG-3′
hu Fascin1	Forward	5′-CCAACGAGAGGAACGTGTCCA-3′
	Reverse	5′-GGAAGGTCTCCTGGTCGGTC-3′
hu IFN-β	Forward	5′-GATTCATCTAGCACTGGCTGG-3′
	Reverse	5′-CTTCAGGTAATGCAGAATCC-3′
hu RIG-I	Forward	5′-TGATAGCAACAGTCAAACACA-3′
	Reverse	5′-AATTTGTCGCTAATCCGTGA-3′
hu MDA5	Forward	5′-TATGAAACCAGAGGAGTATGC-3′
	Reverse	5′-GGTTATTCTTGTAATGCTTGGC-3′
hu IRF-7	Forward	5′-TGCCTCGGAACTGTGAC-3′
	Reverse	5′-CAGGTAGATGGTATAGCGTG-3′
hu IFIT1 (ISG56)	Forward	5′-CGCTATAGAATGGAGTGTCCA-3′
	Reverse	5′-TTTCCTCCACACTTCAGCA-3′
hu IFIT2 (ISG54)	Forward	5′-CTAAAGCACCTCAAAGGGCA-3′
	Reverse	5′-GCATAGTTTCCCCAGGTGAC-3′
hu IFIT3 (ISG60)	Forward	5′-AGTCTAGTCACTTGGGGAAAC-3′
	Reverse	5′-ATAAATCTGAGCATCTGAGAGTC-3′
hu CXCL10 (IP-10)	Forward	5′-TGACTCTAAGTGGCATTCAAGGAG-3′
	Reverse	5′-TTTTTCTAAAGACCTTGGATTAACAGG-3′
TBK1	Forward	5′-TGATAGCAACAGTCAAACACA-3′
	Reverse	5′-TAGAACTTGGATAAATTGGG-3′
IKKϵ	Forward	5′-ACTCTGGAAGTGGCAAGGACAT-3′
	Reverse	5′-TACCTGATCCCGGCTCTTCACCA-3′

Flow cytometric analysis

Apoptotic cells were analyzed by staining with annexin V Alexa 647 and 7-aminoactinomycin D in annexin V staining buffer (10 mm HEPES-KOH, pH 7.4, 140 mm NaCl, 2.5 mm CaCl₂) using a BD FACSCanto II (Becton Dickinson) with Kaluza software (Beckman Coulter, Inc., Brea, CA).

Immunoblot analysis

The cells were washed with Hanks' buffered salt solution and lysed with whole-cell extract buffer (20 mm Tris-HCl, pH 7.4, 150 mm NaCl, 0.5% sodium deoxycholate, 1% NP-40, 1% SDS, 2 mm EDTA, 1 mm PMSF, 250 μg/ml Na₃VO₄, 25 mm NaF, 1× protease inhibitor mixture; Roche Diagnostics, Basel, Switzerland) and then sonicated on ice. Extracts were separated using SDS-PAGE.

Native-PAGE

The cells were lysed with clear native-PAGE cell lysis buffer (50 mm Tris-HCl, pH 7.5, 150 mm NaCl, 1% NP-40, 1 mm PMSF, 5 mm Na₃VO₄, 1× complete protease inhibitor mixture) for 45 min on ice and then centrifuged at 13,000 rpm for 10 min at 4 °C. The protein concentration was adjusted using the Quick Start Bradford protein assay kit (Bio-Rad). Proteins were separated in the absence of SDS in 4–20% gradient gels (Mini-PROTEAN TGX Precast gel; Bio-Rad) using cathode buffer (25 mm Tris, 192 mm glycine, 0.2% sodium deoxycholate) and anode buffer (25 mm Tris, 192 mm glycine). The proteins were transferred to polyvinylidene difluoride membranes, which were probed with a 1:1000 dilution of anti-IRF-3 antibody. Goat anti-rabbit IgG-HRP at a 1:5000 dilution was used as the secondary antibody.

Luciferase reporter assay

HEK293T cells were seeded into 12-well culture plates and transfected with plasmids encoding the −125 + 19 mouse IFN-β luciferase reporter (firefly luciferase; 100 ng) and pRL-TK (Renilla luciferase plasmid; 10 ng, Promega), together with 100 ng of pcDNA3 vector encoding RIG-I, IPS-1, TBK1, IKKϵ, or IRF-3(5D) at increasing concentrations (0, 100, or 200 ng) of pFLAG-CMV4 with fascin1 using GeneJuice transfection reagent (Merck Millipore). IRF-3(5D) is a variant of IRF3 in which Ser-396, Ser-398, Ser-402, Thr-404, and Ser-405 have been substituted with phosphomimetic Asp to produce a constitutively active form (46). Empty pcDNA3 and pFLAG-CMV4 vectors were used to maintain equal amounts of DNA among the wells. The cells were lysed with 1× Glo cell lysis buffer (Promega) 24 h after transfection. Luciferase activity was measured using a dual-luciferase assay (Promega) with a POWERSCAN4 plate reader (DS Pharma Biomedical Co., Ltd., Osaka, Japan) according to the manufacturer's protocol. The reporter gene activity was determined by normalization of the firefly luciferase activity to that of Renilla luciferase.

Co-immunoprecipitation

TBK1, IKKϵ, or IRF-3(5D) were inserted into pFLAG-CMV4 vectors, and fascin1 was inserted into the Myc-pcDNA3 vector for transient transfection into HEK293T cells. Lysates in NP-40 lysis buffer (50 mm Tris-HCl, pH 7.4, 150 mm NaCl, 1 mm MgCl₂, 0.5% NP-40, 10 mm NaF, 1 mm Na₃VO₄, 1 mm PMSF, and 1× protease inhibitor mixture) were incubated with the indicated antibodies for 1 h at 4 °C, followed by the addition of protein G–Sepharose beads (Roche) for an additional 2 h at 4 °C. After centrifugation, the beads were washed with NP-40 lysis buffer and then heated to 98 °C for 5 min in Laemmli sample buffer. Following a brief centrifugation, the supernatants were separated using SDS-PAGE.

Statistical analysis

The statistical significance of the results was calculated using the Student t test or a one-way analysis of variance followed by Tukey's and Dunnett's multiple comparison tests. The differences were considered significant at p < 0.05. All the data were presented as the means ± standard deviation.

Author contributions

T. M., S. H., and S.-I. T. conceived the study, performed the experiments and wrote the paper. M. K. designed and constructed the vectors for the expression of the shRNA and mutant proteins and analyzed the mutant phenotypes in bacteria. C. F. designed, performed, and analyzed the experiments shown in Fig. 3 (Native-PAGE). M. T. provided technical assistance and contributed to the preparation of the figures. A. K. and S.-I. M. analyzed the experiments, coordinated the study, and finalized the paper. All the authors have reviewed the results and approved the final version of the manuscript.