Tankyrase-Binding Protein TNKS1BP1 Regulates Actin Cytoskeleton Rearrangement and Cancer Cell Invasion

Invasion is a dynamic process that involves migration of cells from their original location into depth of the tissue or outside to disseminate to other organs. Enhanced cell invasion is linked to cancer metastasis, the most prominent cause of the intractability of the disease (1). Cell invasion essentially depends on the mechanistic motility of the cell, which is regulated by interactions and signaling from large macromolecular complexes called focal adhesions to the extracellular matrix (ECM). The cellular interface of focal adhesions consists of integrin-α/β heterodimers that bind ECM proteins (e.g., fibronectin, laminin, and vitronectin) and adaptor complexes (e.g., talin, vinculin, and tensin) via the extracellular and intracellular domains, respectively (2). The adaptor complexes capture the retrograde flow of actin filaments (F-actin), and this interaction array of ECM, integrin, adaptors, and F-actin generates tractive force for cell motility (3).

The Rho-associated protein kinases/LIM kinases/cofilin pathway (ROCK/LIMK/cofilin pathway) and CapZ-mediated regulation of actin filament dynamics play key roles in the actin/cytoskeleton network rearrangement (4, 5). ROCKs are serine/threonine kinases that promote actin organization through phosphorylating several downstream targets, including LIMKs (6). Phosphorylated LIMKs then phosphorylate actin-depolymerizing factor/cofilin on serine 3. While cofilin facilitates actin depolymerization, phosphorylation of cofilin on serine 3 (p-cofilin) attenuates its actin depolymerization activity and causes increased numbers of focal adhesion complexes, actin stress fiber formation, and enhanced cell motility (7, 8). Aberrant promotion of LIMK signaling (e.g., by increased expression of the upstream regulators RhoA and ROCK) is observed in many cancers and is associated with cancer metastasis (9, 10). Therefore, LIMK inhibitors, which inhibit generation of p-cofilin, are thought to be promising anti-invasive agents (11).

Tankyrase is a member of the PARP family that catalyzes formation of long PAR chains onto acceptor proteins using NAD⁺ (12). PARylation confers a drastic negative charge to the acceptor proteins and modulates their functions (13). Tankyrase PARylates the telomeric protein TRF1, which is a negative regulator of telomere elongation (12). PARylated TRF1 dissociates from telomeres and is degraded by the ubiquitin/proteasome system. The resulting telomeres exhibit an “open” state that allows easier access of telomerase, which in turn elongates telomeres (14, 15). Tankyrase also upregulates Wnt/β-catenin signaling by PARylation and subsequent degradation of Axins, which are members of the β-catenin destruction complex that consists of Axins, adenomatous polyposis coli (APC), and glycogen synthase kinase 3 (GSK3β; ref. 16). Tankyrase inhibitors stabilize Axins, which in turn promote β-catenin degradation and inhibit the growth of β-catenin–dependent colorectal cancer cells (16, 17). Given that tankyrase has a large protein–protein interaction platform, called ANK repeat clusters (ARC; refs. 18, 19), and is broadly distributed to various intracellular loci, including telomeres, nucleoplasm, nuclear pore complexes, cytoplasm, Golgi, and spindle poles, tankyrase likely possesses yet unidentified functions.

TNKS1BP1 (also called as TAB182) is a tankyrase-binding protein that was identified by a yeast two-hybrid screen (18). This filament-like protein binds to the ARCs of tankyrase and colocalizes with the cortical actin network. However, its biologic function has remained uncharacterized. Here, we demonstrate that TNKS1BP1 interacts with the actin-capping proteins and plays a role in cell motility and invasion. Our observations that TNKS1BP1 depletion facilitates F-actin dynamics and cell invasion through ROCK/LIMK–dependent cofilin phosphorylation establish TNKS1BP1 as a negative regulator of cell motility and invasion. Furthermore, tankyrase also modulates cofilin phosphorylation and cell invasion in a PARP activity–dependent manner, implicating PARylation as a novel posttranslational modulator of cell motility and invasion.

HTC75 cells derived from HT1080 fibrosarcoma cells were obtained from Dr. Susan Smith (New York University School of Medicine, New York, NY) in 2001. PANC-1 and KLM-1 cells were provided by Cell Resource Center for Biomedical Research Institute of Development, Aging and Cancer, Tohoku University (Sendai, Japan) in 2009 and RIKEN BioResource Center in 2010, respectively. They were grown in DMEM supplemented with 10% heat-inactivated calf serum and 100 μg/mL of kanamycin at 37°C in a humidified atmosphere of 5% CO₂. HTC75 cells were authenticated by Seimiya laboratory: these cells contain an exogenous hygromycin-resistant gene and therefore exhibit hygromycin-resistant growth, which has been routinely tested by cultivating the cells with the medium containing 200 μg/mL hygromycin for a week. The growth of HTC75 cells was not affected by this drug treatment, and we used the drug-resistant cells for further experiments in the absence of hygromycin. PANC-1 and KLM-1 cells were authenticated by short tandem repeat profiling analysis (BEX) in 2016.

The detailed information about the expression vectors and antibodies used in this study is given in Supplementary Materials and Methods.

TNKS1BP1 and CapZA2 Stealth siRNAs were purchased from Invitrogen, Life Technologies: TNKS1BP1, 5′-UAUCCAAGCGCUCUUCCCAAACUCC-3′ (#1) and 5′-AAGACGAGGAGUAAUCUUCACCCUG-3′ (#2); and CapZA2, 5′-GCAGCCCAUGCAUUUGCACAGUAUA-3′ (#6). As a control, Stealth RNAi negative control Med GC (#12935-300) was used. Cells were transfected with the siRNAs using Lipofectamine RNAiMAX (Invitrogen, Life Technologies).

Western blot analysis was performed as described (18, 20). Cell lysates were separated by SDS-PAGE, blotted onto polyvinylidene difluoride membranes, and subjected to Western blot analysis with the primary antibodies listed in Supplementary Materials and Methods.

Invasion assay was performed using CytoSelect 96-well Collagen Cell Invasion Assay Kit (Cell Biolabs) according to the manufacturer's instruction. The detailed procedure is given in Supplementary Materials and Methods.

FLAG-tagged TNKS1BP1 was expressed in HEK293T cells, and the cell lysate was immunoprecipitated with FLAG antibody. The immunoprecipitated proteins were analyzed by a direct nanoflow liquid chromatography/tandem mass spectrometry system, as described previously (21).

Subcellular fractions (cytosolic, membrane/organelle, nuclear, and cytoskeletal fractions) were obtained using a ProteoExtract Subcellular Proteome Extraction Kit (Merck Millipore) according to the manufacturer's instruction. Purity of the fractions was confirmed by Western blot analysis with maker proteins: calpain I for cytosolic, histone H2AX for nuclear, and vimentin for cytoskeletal fractions.

Cells were washed with ice-cold PBS and lysed in TNE buffer, containing 10 mmol/L Tris-HCl, pH 7.8, 1% NP-40, 150 mmol/L NaCl, 1 mmol/L EDTA, and 1 mmol/L phenylmethylsulfonyl fluoride, on ice for 30 minutes. Cell lysates were collected after centrifugation at 20,400 × g for 10 minutes at 4°C. Immunoprecipitation was performed as previously described (20). The detailed description is given in Supplementary Materials and Methods.

HTC75 and PANC-1 cells expressing mWasabi-actin (Allele Biotechnology & Pharmaceuticals) were cultured on a poly-l-lysine–coated 35-mm glass-bottom dish (Matsunami Glass) and transfected with siRNAs for 48 hours. To monitor fluorescence recovery after photobleaching (FRAP), we used time-lapse microscopy with a confocal laser scanning microscope (Fluoview FV-1000, Olympus) equipped with a Plan Apochromat 60× oil objective lens (Olympus) and an incubation chamber to ensure a controlled atmosphere (37°C, 5% CO₂). To analyze fluorescence recovery, cells were photobleached using a scanner (100% 488-nm laser transmission, 50 ms) and imaged with a 488-nm laser for excitation and a 510-nm bandpass filter for emission. Three frames were taken every 2 seconds before the photobleaching and then 60 frames (for HTC75) and 40 frames (for PANC-1) were taken every 2 seconds. The fluorescence intensity in the bleached area was measured using ImageJ software (version 1.40g, NIH, Bethesda, MD).

Human tumor tissue microarrays (401-2206, pancreas tumor, matched normal tissue, and pancreatitis) were purchased from Provitro. A pathologist (T. Migita) assessed the data to determine the expression level of TNKS1BP1. Formalin-fixed, paraffin-embedded tissues were collected from 48 pancreatic tumors, some of which contained noninvasive lesions such as intraepithelial neoplasia [pancreatic intraepithelial neoplasia (PanIN), n = 17], under informed consent at the Cancer Institute Hospital, JFCR (Tokyo, Japan). The tissue samples were chosen by two of the pathologists (M. Katori and Y. Ishikawa). Ethical clearance was obtained in advance from the Institutional Review Board of JFCR. The detailed analysis procedure is given in Supplementary Materials and Methods.

In vitro PARP assay was performed as previously described (19, 22). The detailed description is given in Supplementary Materials and Methods.

All data are representative of at least 3 independent experiments. Statistical analysis was carried out using the Student t test.

TNKS1BP1 colocalizes with actin filaments (18; Fig. 1A). Because dynamic remodeling of actin filaments is involved in cellular motility and invasion, we first examined whether TNKS1BP1 affects cellular invasion. As shown in Fig. 1B and Supplementary Fig. S1A and S1B, siRNA-induced TNKS1BP1 depletion in HTC75 fibrosarcoma and PANC-1 and KLM-1 pancreatic cancer cells increased the invasion activity. As a control, cytochalasin D, which disrupts actin polymerization, decreased invasion activity, confirming that this activity depends on actin filament dynamics. In contrast, overexpression of TNKS1BP1 significantly decreased invasion activity (Fig. 1C). These results suggest that TNKS1BP1 affects actin filaments and represses cell invasion.

Cell invasion is promoted by the focal adhesion, a distinct site of cellular adhesion to the ECM (23). Because integrin-mediated tethering of actin filaments forms the focal adhesion, we next addressed whether TNKS1BP1 affects the focal adhesion. Figure 1D shows immunofluorescent staining with paxillin, a marker of focal adhesion, coupled with phalloidin staining, which detects the actin stress fibers (24). Focal adhesions were detected as bright foci at the tips of the actin filaments. TNKS1BP1 depletion increased the number of these foci and the signal intensity of the actin filaments, as compared with the control siRNA–treated HTC75 and PANC-1 cells (Fig. 1D and E and Supplementary Fig. S1C). These observations suggest that TNKS1BP1 negatively regulates cell invasion by repressing the effective assembly of focal adhesion complexes and actin stress fiber formation.

To elucidate the mechanism for the enhanced invasion activity of TNKS1BP1-depleted cells, we monitored the level of p-cofilin. While cofilin facilitates actin depolymerization, p-cofilin attenuates actin depolymerization activity and causes increased number of focal adhesion complexes, actin stress fiber formation, and enhanced cell motility (7). As expected, TNKS1BP1 depletion increased the level of p-cofilin compared with the control cells (Fig. 2A, lanes 1–3 and Supplementary Figs. S1A and S2A). This phenomenon occurred irrespective of the existence of EGF, an inducer of cofilin phosphorylation (Supplementary Fig. S2B; ref. 25).

Phosphorylation of cofilin on serine 3 is mediated by ROCK and its downstream effector kinase LIMK (26, 27). To determine whether the ROCK/LIMK pathway contributes to the increased p-cofilin level in TNKS1BP1-depleted HTC75 cells, we used Y27632, a ROCK inhibitor. As expected, Y27632 attenuated the level of p-cofilin upregulation upon TNKS1BP1 depletion (Fig. 2A, lanes 4–6). As controls, this attenuation was not observed when the cells were treated with either LY294002 (an inhibitor of PI3K) or U0126 [an inhibitor of MAPK/ERK 1 and 2 (MEK1 and 2); Fig. 2A, lanes 7–12]. To further confirm the involvement of LIMK in p-cofilin upregulation, we used a specific LIMK inhibitor (LIMKi). As predicted, LIMKi attenuated the increase in p-cofilin level in TNKS1BP1-depleted HTC75 cells (Fig. 2B). Similarly, Y27632 and LIMKi also repressed p-cofilin upregulation in TNKS1BP1-depleted PANC-1 cells (Supplementary Fig. S3A). These observations indicate that TNKS1BP1 depletion upregulates p-cofilin level through the ROCK/LIMK pathway, resulting in a reduced rate of actin depolymerization.

To examine whether TNKS1BP1 affects the dynamics of actin filaments, we performed FRAP assay of the fluorescent mWasabi-actin–transfected HTC75 cells. In control siRNA–transfected cells, it took 130 seconds to recover 80% of mWasabi-actin signal after photobleaching (Fig. 3). In contrast, in TNKS1BP1-depleted cells, it took only 70 seconds to recover 80% of the signal, which was completely recovered by 120 seconds. TNKS1BP1 depletion in PANC-1 cells gave similar results (Supplementary Fig. S3B). TNKS1BP1-depleted cells exhibited a greater number of actin filaments and dynamic ruffling of the actin-containing lamellipodia (Fig. 3 and Supplementary Videos S1 and S2). These observations indicate that TNKS1BP1 depletion activates the ROCK/LIMK pathway, which induces an inhibitory phosphorylation of cofilin and promotes F-actin dynamics.

To identify the factor that binds TNKS1BP1 and regulates actin reorganization, we performed liquid chromatography/mass spectrometry analysis of anti-FLAG immunocomplexes from the lysates of FLAG-tagged TNKS1BP1–expressing cells. The identified proteins included the actin-capping protein subunits, such as CapZA1, CapZA2, and CapZB, which regulate assembly and dynamics of actin filaments (5). Coimmunoprecipitation assays confirmed that endogenous TNKS1BP1 interacts with all of the ectopically expressed capping proteins in intact cells (Fig. 4A). To examine whether the endogenous capping proteins interact with TNKS1BP1, HTC75 cell lysates were immunoprecipitated with TNKS1BP1 antibody and subjected to Western blot analysis with each anti-CapZ antibody. Figure 4B shows that only CapZA2 was coimmunoprecipitated with TNKS1BP1. Therefore, we focused on CapZA2 in the subsequent analyses.

Coimmunoprecipitation assays showed that CapZA2 directly binds to a C-terminal region (amino acids 1,543–1,635) of TNKS1BP1 (Fig. 4C). To further investigate the relationship between TNKS1BP1 and CapZA2, we fractionated the cells into cytosol, membrane/organelle, nucleus, and cytoskeleton fractions. While CapZA2 was enriched in the cytosol and membrane/organelle fractions, a small amount of the protein was also detected in the nucleus and cytoskeleton in control siRNA–treated cells (Fig. 5A, lanes 1–4). Surprisingly, cytoskeletal CapZA2 disappeared in TNKS1BP1-depleted cells to the equivalent extent as in CapZA2-depleted cells (Fig. 5A, lanes 4, 8, 12, 16). Upon TNKS1BP1 depletion, CapZA2 seemed to be released from the cytoskeleton, as the levels of cytosolic CapZA2 increased in these cells. Elevated p-cofilin was observed not only in TNKS1BP1-depleted cells but also in CapZA2-depleted cells (Fig. 5A, lanes 1, 5, 9, 13 and Supplementary Fig. S1A). These results suggest that TNKS1BP1 works as a scaffold protein that stabilizes CapZA2 in the cytoskeletal F-actin. Meanwhile, TNKS1BP1 was highly sensitive to the cytosolic preparation and extracted into the cytosolic, but not cytoskeletal, fraction. Supplementary Figure S4A confirmed the interaction of TNKS1BP1 with CapZA2 even in a subcellular fraction.

CapZA2-depleted cells showed significantly increased invasion activity (Fig. 5B). Because we did not detect any difference in cell growth among control, TNKS1BP1, and CapZA2 siRNA–treated cells, it is unlikely that the enhanced invasion activity was derived from the altered proliferative potential (Supplementary Fig. S4B). These observations indicate that TNKS1BP1 depletion leads to CapZA2 release from the cytoskeletal actin filaments, which results in enhanced cell invasion.

TNKS1BP1 binds to tankyrase via the 6–amino acid sequence motif (RPQPDG) in TNKS1BP1 (28) with ARCs in tankyrase (18, 19). In agreement with our previous report (18), glutathione S-transferase (GST)-fused TNKS1BP1 C-terminal fragment (GST-TNKS1BP1C) was PARylated by tankyrase, similarly to the positive control GST-TRF1 (Supplementary Fig. S5A, lanes 1–6) but not by the PARP-dead mutant (H1184A/E1291A: HE/A; lanes 7–9). To examine the role of the interaction between TNKS1BP1 and tankyrase, we first performed immunofluorescent staining of TNKS1BP1-depleted cells to assess the PARP activity of tankyrase in an intact cell using TRF1 as a marker. TRF1 is stripped out from telomeres by tankyrase-mediated PARylation and subsequently degraded by the ubiquitin/proteasomal system (14, 15). TNKS1BP1 depletion did not enhance endogenous tankyrase activity, as assessed by the intensities of nuclear (telomeric) TRF1 dots (Supplementary Fig. S5B). Moreover, while overexpression of tankyrase in the nucleus [FN-tankyrase; tagged with FLAG and a nuclear localization signal (NLS) at the amino terminus] diminished TRF1 dots in the nucleus of the control cells, TNKS1BP1 depletion did not change this activity (Supplementary Fig. S5C). Thus, tankyrase PARP activity is not affected by TNKS1BP1 depletion.

Next, we examined the effects of tankyrase overexpression on TNKS1BP1 function. We overexpressed tankyrase either in the nucleus (FN-tankyrase: FLAG-tagged and an NLS at the amino terminus) or in the cytoplasm (F-tankyrase: FLAG-tagged at the amino terminus; ref. 29). These tags did not significantly affect TNKS1BP1 protein level (Fig. 6A). Importantly, the level of p-cofilin was increased when tankyrase was overexpressed in the cytoplasm where actin filaments polymerize (Fig. 6A). To confirm the importance of the cytoplasmic localization and the PARP activity, we used tankyrase (HE/A)-overexpressing cells (29, 30). Only the cytoplasmic and catalytically active tankyrase, but not the nuclear or the HE/A mutant, decreased cytoskeletal protein level of CapZA2 and increased p-cofilin (Fig. 6B). To further confirm the dependency of this phenotype on PARP activity of tankyrase, we examined the effect of the tankyrase PARP inhibitor XAV939 (16). As expected, XAV939 attenuated the elevated level of p-cofilin led by tankyrase overexpression in the cytoplasm (Fig. 6C). Furthermore, only cells that overexpress wild-type tankyrase in the cytoplasm showed significantly enhanced invasion activity (Fig. 6D), which was inhibited by XAV939 (Supplementary Fig. S5D). Neither the HE/A mutant nor overexpression in the nucleus enhanced cell invasion, coincident with Fig. 6B. These observations suggest that tankyrase in the cytoplasm attenuates the function of TNKS1BP1 in a PARP activity–dependent manner to stabilize CapZA2 in the cytoskeletal F-actin, leading to enhanced invasion through rearrangement of the actin cytoskeleton.

We next examined TNKS1BP1 expression in clinical cancers by immunohistochemistry of various tissue microarrays. We found that pancreatic cancers often show reduced levels of TNKS1BP1 protein expression (Fig. 7A and B). Thus, we focused on pancreatic cancer, which is highly aggressive and metastatic with a low survival rate (31). PanIN is a well-defined, common precursor of invasive pancreatic ductal adenocarcinoma arising in the pancreatic duct. We selected patient tissue samples in which we were able to find normal pancreatic ducts, PanINs, and invasive adenocarcinoma in a single field, for easy evaluation of TNKS1BP1 expression changes in the process of cell invasion (32). The intensity of TNKS1BP1 expression remains relatively high in PanINs (Fig. 7C, b) compared with the normal pancreatic ducts (Fig. 7C, a). In contrast, TNKS1BP1 expression was lower in the invasive adenocarcinoma than in the normal pancreatic ducts and PanINs (Fig. 7C, c). Using 17 patient tissue samples, we found that the expression levels of TNKS1BP1 in the invasive regions were significantly lower than those in noninvasive regions of the same tissue (Fig. 7D). These observations indicate an inverse relationship between TNKS1BP1 expression and cancer invasion.

While tankyrase targets TRF1 and enhances telomere elongation by telomerase (14), TNKS1BP1 does not bind telomeres or affect telomere length in human cancer cells (ref. 33 and H. Seimiya, unpublished observations). MOTIF analysis (http://www.genome.jp/tools/motif/) indicates that there are no characteristic motifs or known functional domains in TNKS1BP1. According to the FASTA homology search (http://www.genome.jp/tools/fasta/), however, this protein has a weak similarity to ECM proteins, such as collagens and proteoglycans, suggesting a role in cell adhesion, motility, and/or invasion. This information and its intracellular colocalization with actin filaments support that TNKS1BP1 is a functional component of the actin cytoskeleton.

In our original report, we also detected TNKS1BP1 in the nucleus, especially in the perinucleolar heterochromatic regions (18). We speculate that these nuclear TNKS1BP1 would be derived from alternative splicing, different start sites, or both of the transcripts as observed in the H-InvDB Annotated Human Gene Database (http://www.h-invitational.jp), which give rise to shorter forms (i.e., the C-terminal half of full-length protein). Consistent with this idea, the TNKS1BP1-truncated mutant that lacks the N-terminal 1220 amino acids accumulates in the nucleus when ectopically expressed in cells (H. Seimiya, unpublished observation). Although TNKS1BP1 contains 2 NLS at the C-terminal basic region, their function appears to be recessive in the full-length protein because the full-length TNKS1BP1 colocalizes with actin filaments rather than in the nucleus. Nuclear localization of the shorter isoforms is further supported by the fact that antibodies raised against the N-terminal fragment of TNKS1BP1 detect only cytoskeletal TNKS1BP1 but not nuclear TNKS1BP1 by immunofluorescent staining (H. Seimiya, unpublished observation). These facts and functional analyses in the present study establish full-length TNKS1BP1 as a cytoskeletal protein. Although the function of the short isoforms in the nucleus remains to be determined, Zou and colleagues demonstrated that TNKS1BP1 functions in DNA double-strand break repair (34).

We showed that TNKS1BP1 expression level is a determinant for the actin cytoskeletal rearrangement and the ability of the cell to invade into the collagen matrix. Intriguingly, TNKS1BP1 depletion decreased the level of CapZA2, a newly identified TNKS1BP1-binding partner, in the cytoskeletal fraction. CapZA2 is a component of the heterodimeric actin-capping protein, which consists of α (CapZA1 or CapZA2) and β (CapZB) subunits (35). They cap the barbed ends of growing actin filaments to block their elongation (5). This effect is similar to those of cytochalasins, which bind the barbed ends of growing actin filaments and inhibit cell invasion (ref. 36 and this study). Our results suggest that TNKS1BP1 stabilizes the capping proteins at barbed ends and negatively regulates actin polymerization.

What, then, is the molecular mechanism for CapZA2 dissociation from actin filaments upon TNKS1BP1 depletion? Because there is a significant mismatch, that is, different symmetries between 3-dimensional structures of CapZ and actin (37), it is possible that TNKS1BP1 is required for efficient interaction between these two proteins. CapZ interaction with actin is also inhibited by phosphatidylinositol 4,5-bisphosphate, myotrophin/V-1 and CARMIL (38–40). Whether TNKS1BP1 knockdown may affect the intracellular levels, distributions, or both of these factors remains to be investigated.

We found that knockdown of either TNKS1BP1 or CapZA2 increases the level of p-cofilin in a ROCK/LIMK pathway–dependent manner, leading to an increased number of focal adhesions that were accompanied by F-actin polymerization. So far, the precise mechanism underlying why CapZA2 dissociation from the actin filaments activates the ROCK/LIMK pathway remains elusive. Given that ROCK inhibition enhances the protein levels of CapZ (41) and strain-induced stimulation of CapZ dynamics depends on the RhoA/ROCK pathway (42), there might be a feedback loop system in the RhoA/ROCK/LIMK pathway and the CapZ-mediated regulation machinery of actin reorganization.

We demonstrated that TNKS1BP1 is downregulated in invasive pancreatic adenocarcinoma. This observation suggests that TNKS1BP1 may act as a suppressor of pancreatic cancer invasion via negative regulation of actin cytoskeleton rearrangement. Given that TNKS1BP1 downregulation is observed exclusively at the invasive area of the cancerous lesions, it is possible that its protein expression is regulated in a microenvironment-dependent, transient manner. In fact, the Oncomine database (https://www.oncomine.org) shows that the level of TNKS1BP1 transcript is not significantly altered in the bulk of pancreatic cancer cells. This would be reminiscent of the epithelial–mesenchymal transition (EMT), through which an epithelial cancer cell converts to a mesenchymal cell type with less cell-to-cell adhesion and higher motility (43). Microenvironmental factors, such as TGFβ, promote EMT of metastasizing cancer cells. Furthermore, it has been postulated that cancer cells having settled at the metastatic site often undergo the reverse event, mesenchymal–epithelial transition (MET). Thus, EMT and MET are mutually reversible, and the balance between these two events regulates cell motility and invasion. We monitored epithelial and mesenchymal marker proteins, such as E-cadherin and N-cadherin, respectively, and found no evidence that TNKS1BP1 directly regulates EMT (T. Ohishi & H. Seimiya, unpublished observation).

Our data suggest that TNKS1BP1 regulates CapZ recruitment to actin cytoskeleton. Recently, Lee and colleagues reported that CapZA1 expression is associated with decreased cancer cell migration and invasion and could be used as a good prognostic marker of gastric cancer (44). The authors also showed that CapZA1 depletion causes a significant increase in gastric cancer cell migration and invasion, whereas CapZA1 overexpression shows the opposite effects. These results are in good agreement with our observation that CapZA2 depletion increases the invasive activity of cancer cells. Together, these observations suggest that TNKS1BP1-CapZ interaction with the actin cytoskeleton plays a negative role in cancer cell invasion.

Our result that tankyrase-overexpressing cells phenocopy TNKS1BP1-depleted cells in terms of (i) CapZA2 release from actin filament, (ii) p-cofilin upregulation, and (iii) enhanced cell invasion suggest that tankyrase is the upstream repressor of TNKS1BP1 function. These effects of tankyrase require its PARP activity and cytoplasmic localization. According to Tian and colleague, tankyrase inhibition by a small compound or by RNA interference reduces the invasive activity of neuroblastoma cells, depending on telomere shortening (45). Because the effect of tankyrase inhibition on telomere function emerges rather slowly due to gradual shortening of telomeres (33, 46), we prefer a non-telomeric mechanism for the altered invasive activity.

Tankyrase-mediated PARylation of the Wnt suppressor Axins leads to ubiquitination of the PARylated Axins followed by proteasomal degradation (16). This is a unique signaling relay from tankyrase-mediated PARylation to RNF146-mediated ubiquitination: the ubiquitin E3 ligase RNF146 in its resting state binds tankyrase, and PAR chains produced by tankyrase bind the WWE domain of RNF146, which causes allosteric activation of the ubiquitin E3 ligase activity (47). Tankyrase-mediated PARylation followed by ubiquitination is also observed in other tankyrase substrates, such as TRF1 and 3BP2 (15, 48). In contrast, tankyrase inhibitors did not upregulate the level of TNKS1BP1 protein (T. Ohishi & H. Seimiya, unpublished observation), suggesting that PARylated TNKS1BP1 may not be an efficient substrate for ubiquitin-dependent proteasomal degradation.

The question regarding the mechanism by which tankyrase regulates the function of TNKS1BP1 remains unanswered. Tankyrase has 5 ARCs as a platform for protein–protein interactions, and ARCs 1, 2, 4, and 5 can bind TNKS1BP1 (19, 49). Existence of multiple ARCs suggests a role for tankyrase as a molecular lattice, in which TNKS1BP1 may be an essential component for functional linkage to the CapZ actin filaments. Formation of these protein complexes would depend on the specific interaction between tankyrase and TNKS1BP1 because TNKS1BP1 binds tankyrase ARCs but not the conventional ankyrin G (18). Consistent with this idea, efficient Wnt/β-catenin signaling requires the lattice-like scaffolding of tankyrase, which is mediated by the sterile α motif–dependent polymerization and multivalent ARC interaction with Axins (50).

In conclusion, we have shown that TNKS1BP1 interacts with actin-capping proteins and regulates the ROCK/LIMK/cofilin pathway for actin reorganization. These observations give a new insight into the molecular mechanism for actin-regulated cell motility and how its perturbation could contribute to cancer invasion.

No potential conflicts of interest were disclosed.

Conception and design: T. Ohishi, H. Seimiya

Development of methodology: T. Migita, H. Seimiya

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): T. Ohishi, H. Yoshida, M. Katori, Y. Muramatsu, M. Miyake, Y. Ishikawa, A. Saiura, S.-i. Iemura, H. Seimiya

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): T. Ohishi, M. Katori, T. Natsume, H. Seimiya

Writing, review, and/or revision of the manuscript: T. Ohishi, H. Seimiya

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): H. Seimiya

Study supervision: H. Seimiya

The authors thank Dr. Sho Isoyama for technical assistance to quantify the immunohistochemical staining, Drs. Kazuhiro Katayama and Yoshikazu Sugimoto for the 3×HA plasmid, and Dr. Toru Hirota for the pcDNA3-FLAG destination vector.

This work was supported in part by a Grant-in-Aid for Young Scientists (B), Japan Society for the Promotion of Science (JSPS; no. 23701068, 25871074 to T. Ohishi), a Grant-in-Aid for Scientific Research on Innovative Areas, Ministry of Education, Culture, Sports, Science and Technology (no. 23117527 to H. Seimiya), and a Grant-in-Aid for Challenging Exploratory Research, JSPS (no. 26640109 to H. Seimiya).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.