CARF enrichment promotes epithelial–mesenchymal transition via Wnt/β-catenin signaling: its clinical relevance and potential as a therapeutic target

CARF (Collaborator of ARF)/CDKN2AIP was discovered as a novel ARF-binding protein. It has been established as an essential cell survival, p53-, and cell proliferation-regulatory protein. Although a moderate upregulation of CARF caused growth arrest and senescence, its excessively enriched levels were shown to facilitate aggressive proliferation and malignant transformation of cancer cells. Here, we examined the relevance of CARF levels in clinical tumors and found its amplification (both at gene and transcript levels) in a variety of invasive and metastatic malignancies. Consistent with the clinical readouts, enrichment of CARF in cancer cells promoted epithelial–mesenchymal transition (EMT). Cancer database and molecular analyses revealed that it activates Wnt/β-catenin signaling axis, as evident by enhanced nuclear localization and function of β-catenin marked by increased level of SNAIL1, SNAIL2, ZEB1, and TWIST1 and its downstream gene targets. Of note, targeted knockdown of CARF led to decrease in nuclear β-catenin and its key downstream effectors, involved in EMT progression. Consistent with this, CARF targeting in vivo either by naked siRNA or CARF shRNA harboring adeno-oncolytic virus caused suppression of tumor progression and lung metastasis. Taken together, we report clinical and therapeutic relevance of CARF in EMT and cancer invasiveness/metastasis, and propose it as a potent therapeutic target of aggressive cancers.


INTRODUCTION
Systematic diagnosis, screening, and targeted therapeutic approaches have considerably improved clinical response and survival of cancer patients in last decade. Furthermore, molecular profiling of aggressive tumors has not only led to improved prognosis, but also helped to minimize the therapeutic side effects in several common malignancies 1,2 . However, risk of metastases and recurrence of disease remained largely inevitable and unpredictable, and hence warrant continued efforts to characterize the associated underlying mechanisms. Transformed cancer cells progressively loose cell-cell adhesion/epithelial traits and acquire invasive mesenchymal characteristics via a systemic cellular reprograming, viz. epithelial-to-mesenchymal transition (EMT). EMT facilitates spread of cancer from the site of origin to other organs via acquiring motility, activation of extra-cellular matrix (ECM)-degrading proteases enabling invasion and eventual dissemination through vascularization 3 . Conferring reorganization of cytoskeleton and cellular polarity, it plays indispensable role in biological processes including embryonic development and wound healing 4,5 , and has been tightly associated with cancer metastases, drug resistance, and recurrence [6][7][8] .
During the process of EMT, cells progressively loose set of proteins required for cell-adhesion/tight-junction (e.g., E-cadherin, occludin, and ZO-1) and gains others (e.g., vimentin, N-cadherin, or fibronectin) that offer mesenchymal properties 6,9 . Transcription factors, viz. Slug, Snail, and Twist have been demonstrated to regulate this process and are in turn influenced by diverse signaling pathways including TGF-β1, RTK (Receptor Tyrosine Kinase) receptors, i.e., EGFR, ERBB2/HER2-AKT, ITGB1/FAK, NOTCH, and Wnt/β-catenin proteins 4,10 . In particular, deregulated function of TGF-β1, a regulatory cytokine has been associated with tumor initiation and metastases 11 . It is frequently amplified in metastatic breast and pancreatic cancers 12 and found to promote malignant and metastatic characteristics via activation of Snail and Zeb1 13 . Similarly, the RTK receptors, EGFR and ERBB2 (frequently amplified in glioblastoma, breast and esophagus cancers) have been shown to promote EMT 12 . Focal adhesion kinase (FAK) and AKT, serine/threonine kinases are upregulated and shown to promote EMT in various malignancies 12 . Upregulation of Wnt in cancer cells has been shown to stabilize β-catenin and promote its translocation to the nucleus, where it serves as a coactivator of TCF-LEF transcription factors and regulate several EMT regulating proteins including repression of E-cadherin and activation of Snail, Zeb1, and Vimentin 3,10, 13 .
We have previously cloned CARF (the Collaborator of ARF), as a novel interactor of tumor suppressor protein p14 ARF (ARF) 14 . It has been shown that CARF activates Fig. 1 Amplification of CARF and its enrichment in metastatic malignancies. a Percentage CARF amplification in clinical cancer datasets in TCGA database. b Enrichment of CARF mRNA levels in invasive and metastatic malignancies; RNA-Seq (V2 algorithm) expressions in log10 scale. c CARFgene amplification in metastatic breast cancer cell lines (upper); values represented as fold enrichment (lower). d Dot plot showing percentage amplification of candidate genes (shortlisted in TCGA-defined invasion/metastases and angiogenesis gene sets) in patients with (n = 10) and without (n = 97) CARF amplification in NEPC (2016) dataset p53, a key tumor suppressor in a ARF-dependent and independent manners 15,16 leading to execution of growth arrest and pre-mature senescence, as marked by p21 WAF1 activation in human cancer and normal cells, respectively 17,18 . Excessively enriched or super-high expression level of CARF was shown to promote pro-proliferation and malignant properties of cancer cells in a feedback regulatory manner involving p53-HDM2 and DNA damage regulating proteins (Chk1, Chk2, ATM, and ATR) 18 . Most recently, we found that CARF led to transcriptional repression of p21 WAF1 and contributes to malignant transformation remarkably detected in p53deficient cells 19 . In light of above information, in the present report, we investigated clinical relevance of CARF in tumorigenesis and its progression. An initial survey marked amplified CARF levels in invasive and metastatic malignancies in public cancer patient databases and in clinical tumor samples. We generated CARF-enriched cells and performed in vitro and in vivo assays, and found that CARF enrichment leads to EMT progression via Wnt/β-catenin signaling axis as evidently found by βcatenin nuclear translocation and activation of TCF4/βcatenin transcriptional targets. CARF knockdown, on the other hand, abolished these activities both in in vitro and in vivo models demonstrating functional relevance of CARF in EMT progression and its importance as a therapeutic target for metastatic malignancies.

CARF-amplification marks tumor invasiveness and metastases
In order to examine the status of CARF across malignancies, we surveyed its genomic alterations (Copy Number Alterations, CNA) in clinical tumor patient in normal, infiltrating, and metastatic breast cancer patient tumors; their quantitative levels shown on right. b, CARF levels in patients with normal to infiltrating (sub-group #1) and infiltrating to metastatic (sub-group #2) breast tumors in different age groups; quantitation shown below. c Expression of CARF in normal and metastatic tumor sections of intestine, liver and skin; quantitation shown on right. d Quantitation of CARF expression in normal and malignant breast, prostate, skin, liver, and kidney tumors sections; and in normal, lobular, and ductal breast cancer immunohistological sections retrieved from Human Protein Atlas (e) database (TCGA). As shown in Fig. 1a, we found CARF amplification in several tumor types of which prostate cancer (Neuroendocrine; NEPC, Trento/Cornell/Broad 2016 dataset) scored the highest (9.3%). Expression data revealed elevated CARF mRNA levels across datasets in various malignant tumors ( Fig. 1b; Figure S1a). We next crosschecked CARF levels at Oncomine, an online cancer microarray expression database, and found its amplification in invasive breast carcinomas as compared to normal tissue ( Figure S1b). Several other invasive malignancies such as sarcoma, lung, and prostate cancers, showed enrichment of CARF expression at Oncomine, suggesting clinical relevance of CARF upregulation with metastatic cancers. In view of these findings, we examined and found genomic amplification of CARF (2-7 fold) in number of invasive breast cancer cell lines, derived from metastatic sites (Fig. 1c). In order to further confirm CARF correlation with metastasis, we examined the status of 31 genes (pooled from defined invasive/metastatic and angiogenesis gene sets, at TCGA) in NEPC (2016) dataset having patients with (n = 10) and without (n = 97) CARF amplification. As shown in Fig. 1d, we found that patients with CARF amplification showed 40% or higher amplification of 28 out of 31 genes. Three genes (MMP2, MMP15 and CDH1) showed low (<20%) level of amplification. Of note, CDH1, shown to decline during invasion or EMT progression [20][21][22] , showed the least change in CARF-amplified patients. These data strongly suggested that CARF amplification is associated with metastatic properties of the cancer cells. We next examined the expression level of CARF in clinical tumor sections using tissue microarray. Of note, a significant and gradual increase in CARF was detected from progression of infiltrating to metastatic breast carcinoma as compared to the basal level in normal matched tissues (Fig. 2a). Comparison of two patient sub-groups of similar age, bearing normal, infiltrating and metastatic tumors, revealed significant CARF enrichment in the latter stages (Fig. 2b). While a moderate increase in CARF expression was detected in intestine and liver malignant tumors as compared to their matched normal tissues, metastatic melanoma showed abundant enrichment as compared to normal skin (Fig. 2c). We crosschecked it with the Human Protein Atlas and observed enriched CARF expression in breast, prostate, skin, liver and kidney carcinomas as compared to their respective normal tissues (Fig. 2d). Consistent to mRNA expression data ( Figure  S1b), invasive ductal tumors showed a higher level of CARF expression than the lobular breast carcinoma (Fig.  2e). Together with the data in Fig. 1, these findings confirmed the involvement of CARF in cancer metastasis.

CARF enrichment enhanced cancer cell migration and invasiveness
In order to get molecular insights to the role of CARF in EMT and cancer metastasis, we recruited a panel of in vitro cultured cancer cells and examined their level of CARF expression. We found that invasive ovarian (SKOV-3), lung (H-1299), and fibrosarcoma (HT-1080) cell lines derived from secondary metastatic tumor site possessed high level of CARF expression (Fig. 3a). Although the non-metastatic osteosarcoma (U2OS) showed lowest level of CARF expression (Fig. 3a). Analyses of CARF status in cancer cell lines in CCLE (Novartis/Broad 2012) TCGA dataset also revealed CARF-gene amplification and transcript enrichment in number of circulating/blood-born and invasive cancer cell lines derived either from primary or secondary metastatic tumor site ( Figure S1c) suggesting that CARF may have a key role in cancer metastasis.
We next generated CARF-enriched U2OS and SKOV-3 cells, representing low and high endogenous CARF levels, respectively. To distinguish it from endogenous CARF, the ectopic CARF was tagged with GFP ( Fig. 3b). Cell phenotype analyses revealed that CARF enrichment enhanced their mesenchymal characteristics (Fig. 3b). As shown in Figs. 3c and d, the CARF-enriched cells exhibited higher migration and invasiveness. Analysis of several EMT marker proteins in control and CARF-GFP U2OS and SKOV-3 cells revealed their upregulation in the latter ( Fig. 3e; Figure S2a). Consistently, enhanced expression of vimentin, N-cadherin, and fibronectin in U2OS/SKOV-3 CARF-enriched cells confirmed promotion of their mesenchymal characteristics ( Fig. 3f; Figure S2b). It was further affirmed by upregulated MMP1, MMP3 and MMP8, and hnRNP-K ( Fig. 3g; Figure S2c) and downregulated cytokeratins levels in these cells (Fig. 3h).

CARF enrichment promoted EMT via activating Wnt/βcatenin signaling
Considering the above data that demonstrated the role of CARF enrichment in EMT, we next investigated enrollment of key EMT regulating signaling axes. Canonical factors including TGF-β, tyrosine kinase receptors (i.e., EGFR, HER2/ErbB2), Wnt/β-catenin, and NOTCH1 have been established as the key regulators of EMT 6,23,24 . We next examined their correlation with CARF in TCGA cancer database. In a co-occurrence-based analyses, taking amplification of these key modulators in their top 10 datasets, we found an association of β-catenin amplification with CARF ( Figure S3a). As no consistent association of any of the other regulators, TGFB1, EGFR, ERBB2 and NOTCH1 was found with CARF amplification ( Figure  S3b-e). To test it further, we analyzed status of transcriptional targets of TCF4/β-catenin in NEPC (2016) dataset. A large proportion of these genes were found to be significantly upregulated in above dataset ( Figure S3f). Of note, TCF4/β-catenin direct target genes involved in regulating EMT process showed a significant increase, except of CDH1 i.e. a repressive target ( Figure S4). Furthermore, dataset exhibited a positive association in the expression level of indirect TCF4/β-catenin gene targets categorized as repressor and activators of EMT process ( Figure S4). Gene amplification frequencies in the dataset also revealed an association of CARF with that of DVL2, TCF4, LEF1, and JUN genes suggested to make a complex to stabilize TCF4/β-catenin interaction in the nucleus ( Figure S5a). Besides these, amplification of number of β-catenin gene targets including TDG, SIAH1, CSNK1A1, BTRC, FRAT1, MAP3K7, SMAD4, and TLE1 also found to be associated with CARF ( Figure S5b). These data suggested that CARF might be involved in regulation of Wnt/β-catenin signaling.
To investigate the above correlation in vitro, we examined β-catenin expression in U2OS/CARF-GFP cells and found its remarkable enrichment in the nucleus (Fig.  4a), suggesting its transcriptional activation. The latter was endorsed by upregulation of β-catenin gene targets, SNAI1, SNAI2, ZEB1, and TWIST1 transcription factors, known to be the key instigators of EMT process ( Fig. 4b; Figure S6a). Of note, SNAI1 and TWIST1, two direct targets of β-catenin exhibited greater increase as compared to the SNAI2 and ZEB1, i.e., the indirect targets of β-catenin, in both U2OS and SKOV-3/CARF-GFP cells (Fig. 4b, Figure S6a). Of note, we observed a repression of CDH1 transcript levels in these cells (Fig. 4b, Figure S6a). Transcript levels of a number of gene targets that interact in cytoplasm, e.g., DVL2, AXIN2, and nucleus, e.g., TCF4, LEF1 to form TCF4/β-catenin complex were upregulated in CARF-enriched U2OS (Fig. 4c) and SKOV-3 cells ( Figure S6b). Together, these data suggested that CARF enrichment led to activation of TCF4/β-catenin function and repression of E-cadherin, an epithelial marker; declines during EMT progression. Furthermore, steady expression of TGFβ1, its receptors (i.e., TGFBR1) and of FAK and AKT, the key EMT regulators of TGFβ and Tyrosine/Focal-adhesion-kinase signaling axis, respectively, further ruled out their involvement in CARFpromoted EMT function in these cells (Figures S6c-d).
Consistent to the increase in β-catenin transcript, and its nuclear translocation (Fig. 4a), we found increased  (Figure 4e). β-catenin, Survivin, SMAD2/3, and Cyclin D1 proteins also showed an increase in CARF-enriched U2OS and SKOV-3 cells (Fig. 4f, Figure S6e). Of note, TGFβ1 protein levels remained unchanged in these cells (Fig. 4g), consistent to its transcript levels ( Figure S6c). On the other hand, upregulation and nuclear translocation of SMAD2/3, also a gene target of β-catenin further implies its regulation in CARF-enriched cells (Fig. 4g). These data suggested that CARF regulates nuclear translocation of βcatenin, thereby instigating its transcriptional function. To investigate further the mechanism of CARF-driven βcatenin activation, we considered two possibilities; involvement of (i) Wnt ligand and (ii) kinase pathway. Accordingly, we enrolled IWP2, a specific inhibitor of Wnt ligand 25 and Pyrvinium pamoate (PP), an allosteric activator of Casein Kinase 1α (CK1α) 26 that stimulates ubiquitination and degradation of β-catenin by its phosphorylation 27 . As shown in Fig. 4h, in contrast to the untreated CARF-enriched control U2OS cells, PP-treated cells showed marked decrease in β-catenin protein and its targets (Survivin and Cyclin D1); such decrease was even more evident in CARF-enriched SKOV-3 cells ( Figure  S7a). Similar results were obtained with iCRT14, a potent inhibitor that interrupts β-catenin and TCF4 binding and halts β-catenin-mediated transcription (Fig. 4h, Figure  S7b). On the other hand, treatment of IWP2 appeared not to significantly alter CARF-induced upregulation of βcatenin, Survivin and Cyclin D1 (Fig. 4h). These data suggested that CARF-driven activation of β-catenin function is essentially regulated through kinase pathway. These results were further supported by immunostaining  Fig. 4a and i, CARF-enriched control and IWP2, and not the PP, treated cells showed nuclear translocation of β-catenin. Similar results were obtained in SKOV-3 cells ( Figure S7c).

CARF knockdown attenuated cancer cell invasion and migration via abrogating nuclear translocation of βcatenin
We next investigated whether acquired EMT characteristics in CARF-enriched cells via activation of Wnt/ β-catenin axis could be reversed by CARF suppression. CARF-specific siRNA (20-200 pM; 72 h) mediated suppression of CARF in SKOV-3 cells (higher level of endogenous CARF expression) resulted in reduced survival, a housekeeping CARF function (Fig. 5a) as also reported earlier 28 . We opted to use sub-toxic concentrations (20-50 pM) of CARF siRNA for 48 h in in vitro assay. CARF-compromised SKOV-3 cells showed decrease in migration and invasion properties (Figs. 5b, c). Furthermore, dose-dependent inhibition of CARF siRNA (~10-50 pM, sub-toxic doses) on migration was observed in highly invasive cancer cells (viz. HT-1080 and MDA-MB-231) (Figures S8a and S8b). These cells led to a gradual inhibition of cell migration with increasing CARF siRNA concentration, as observed by cell morphology and activity in wound-healing assay (Figures S8c-S8d and data not shown). Analysis of protein expressions revealed downregulation of number of mesenchymal and invasion marker (Fig. 5d), while E-cadherin levels elevated. Of note, compared to control, CARF-compromised SKOV-3 cells showed a dispersed and reduced nuclear β-catenin accumulation (Fig. 5e). Additionally, CARF-silenced cells possessed higher E-cadherin and reduced vimentin (Fig.  5f), N-cadherin, and fibronectin (Fig. 5g) expression as compared to the control cells.
We next determined the role of Wnt/β-catenin in CARF-induced EMT. As shown in Fig. 6a, CARF silencing led to a sixfold decrease in nuclear β-catenin expression; although the total β-catenin level did not show significant change (Fig. 6a). CARF silencing also led to decrease in Wnt1 nuclear foci in both U2OS and SKOV-3 cells (Fig.  6b). Of note, nuclear levels of survivin and cyclin D1 showed decrease (Fig. 6c). Similarly, these cells showed a reduced c-Myc accumulation in nucleus in both U2OS and SKOV-3 cells (Fig. 6d). We found that CARFcompromised cells possess reduced level of SNAI1 and TWIST1, the direct targets of β-catenin (Fig. 6e). Although SNAI2 and ZEB1 transcript levels altered slightly. These data suggested that CARF silencing led to abrogation of TCF4/β-catenin function marked greater effect on its immediate/direct targets (SNAI2, ZEB1). α-Smooth Muscle Actin (α-SMA/ACTA2), a key EMT marker also showed significant decrease in CARFcompromised cells (Fig. 6e, lower). Analysis of transcript levels of Wnt1 and β-catenin partners in cytoplasm and nuclear complex revealed that along with Wnt1, AXIN2, TCF4, and LEF1 showed a significant decrease, although DVL2 decrease marginally (Fig. 6f). Of note, β-catenin transcript level remained largely unchanged in control and CARF-compromised cells (Fig. 6f). On the other hand, the latter showed nuclear depletion of β-catenin, with its unaltered total protein levels (Figs. 6a, g) suggesting an inhibition of its transcriptional activation function by its nuclear depletion as was endorsed by reduced expression of its gene targets and effector proteins. As shown in Fig. 6g, Wnt1, Survivin, and SMAD2/3 were reduced in CARF-compromised cells. Besides these, we observed nuclear depletion of SMAD2/3 in CARFcompromised cells (Fig. 6h). The results were confirmed in CARF-compromised U2OS cells (Figure S9a-c). Of note, these cells, as in SKOV-3, showed SMAD2/3 depletion from the nucleus (Figure S9d), suggesting an inhibition of its nuclear function, downstream of TCF4/βcatenin activity, critically involved in EMT.

CARF knockdown attenuated tumor metastases in xenograft models
To examine the physiological relevance of CARFpromoted EMT and its abrogation upon its suppression, we performed in vivo xenograft assay. Intravenously and subcutaneously injected nude mice with SKOV-3 cells (control groups), were treated with control and CARF siRNAs, as shown in the treatment regime (Fig. 7a). On day 40, observation of their liver and lungs revealed metastatic foci/sacs in control and CARF-enriched cellinjected group, however, their size was found to be larger in the latter (Fig. 7a). Of note, the group given intermittent injections of CARF siRNA had normal liver and lungs (Fig. 7a). Concomitantly, tumors given intratumoral CARF siRNA injections harbored smaller-size tumors compared to the control, while mice injected with CARFenriched cells raised larger tumors (Fig. 7b). Mice bodyweight and tumor-weight analyses revealed that CARF enrichment promoted aggressive tumors and its knockdown led to suppression of their growth (Fig. 7a, b).
We next employed CARF targeting oncolytic adenovirus with lung metastatic xenograft model using luciferase-expressing invasive lung cell line, i.e., A549. As shown in Fig. 7c, treatment of RdB-shCARF resulted in reduced metastatic progression in lungs. At 21 days post injection, the luciferase signal from the metastatic lung tumors of mice treated with RdB-shCARF (2.4 × 10 7 p/s) was significantly attenuated in comparison to other groups demonstrating 186.2-, or 82.2-fold greater therapeutic efficacy than PBS (4.5 × 109 p/s), or RdB (2.0 × 109 p/s), respectively (Fig. 7c, d), these data demonstrated and endorsed the physiological relevance of CARF in metastases and its inhibition by CARF knockdown.

DISCUSSION
CARF has been established as an essential protein that poses two-way control on cell proliferation by regulation of ARF-p53-HDM2-p21 WAF1 signaling [15][16][17][18] . Overexpression of CARF in ARF-dependent or independent manner was shown activate and stabilize p53 function causing pre-mature senescence 16,17 . On the other hand, excessively enriched/super-high level of CARF expression promoted proliferation and malignant properties of cancer cells 18,19 . In the present study, enrichment of CARF (gene and transcript) levels in cancer patient datasets/ tumor samples suggested clinical relevance of CARF upregulation with cancer metastasis (Figs. 1 and 2). Furthermore, patients with amplified CARF marked higher amplification level of several markers involved in metastasis and angiogenesis (Fig. 1d). In vitro analyses of markers involved in EMT, invasion, and migration 21,22 (Fig. 3e-h; Figure S2) exhibited their upregulation in CARF-enriched both U2OS and SKOV-3 cells. The clinical readouts and molecular analyses conferred that enrichment of CARF is tightly associated with EMT, metastasis and invasion. CARF amplification and enriched expression levels showed an association with β-catenin across tumor datasets ( Figures S3). Of note, CARFenriched cells showed a distinct nuclear enrichment and transcriptional activation of β-catenin as supported by increased expression of its gene targets including SNAIL1, TWIST1, SNAI2, ZEB1, and effectors proteins. By molecular analyses and use of specific inhibitors, we determined that CARF activates Wnt/β-catenin signaling through kinase pathway.
The above data demonstrated a tight correlation and suggested an important role of CARF in EMT progression. siRNA-mediated CARF knockdown in SKOV-3, HT-1080, MDA-MB-231 invasive cells caused inhibition of cell migration and invasive characteristics. Furthermore, reduced mesenchymal (vimentin, N-cadherin, fibronectin) and elevated epithelial (E-cadherin) properties in these cells supported that CARF could be a potential therapeutic target in reversing the EMT process. Of note, whereas CARF enrichment caused nuclear translocation of β-catenin, its abolition was marked in CARFcompromised cells demonstrating that CARF has essential role in activation of β-catenin signaling involved in EMT. The data was endorsed by downregulation of several β-catenin-regulated proteins including, survivin, c-Myc, Cyclin D1, Wnt1, and SMAD2/3. Furthermore, βcatenin interacting proteins required for active transcriptional complex (AXIN2, DVL, TCF4 & LEF1) and effectors (Cyclin D1, c-Myc, survivin) showed increase in CARF-enriched and decrease in CARF-compromised cells.
Recently, Fan et al. 29 , using an oncogenic Ras model of hepatocellular carcinoma (HCC), reported that CARF promotes β-catenin/TCF activation by inhibition of βcatenin and ICAT interaction. He et al. 30 reported a small-molecule inhibitor of CARF. Using in vitro and embryonic developmental model of zebra fish, they showed that suppression of CARF led to impaired embryonic development, hematopoietic and caudal fin regeneration. The study showed a direct CARF-Dvl interaction that facilitated TCF/β-catenin transcriptional activity 30 . In continuation with our earlier studies that demonstrated that CARF is an essential and dual regulator of cell proliferation [14][15][16][17][18][19] . We provide clinical relevance of CARF upregulation in malignant progression of cancer and mechanism of its role in EMT, migration and invasion (Fig. 7e). Of note, frequencies of CARF amplification with that of β-catenin, Dvl2, TCF4, and c-Jun across patient datasets implying a possible association of CARF in stabilizing Dvl-c-Jun-β-catenin-TCF transcriptional complex as was also hypothesized by He et al. 30 Markedly, CARF-induced transcriptional activation of TCF/β-catenin could be reversed by siRNAmediated CARF silencing in in vitro as well as in vivo models suggesting that targeting of CARF has therapeutic potential for treatment of metastatic and invasive cancers.
Taken together, we report that CARF (i) amplification is clinically relevant to cancer invasiveness and metastasis, (ii) facilitates EMT and metastatic progression via activation of Wnt/β-catenin signaling and (iii) is a strong therapeutic target for malignant and aggressive cancers.

Reverse transcription (RT)-PCR, Real-Time PCR, and Genomic PCRs
Semiquantitative RT and Real-Time PCRs were performed as described earlier 18,31 . Genomic PCR was performed on a panel of normal and metastatic breast cancer cell lines (American Type Culture Collection, ATCC) as described earlier 32 using CARF-specific primer sets (Table  S2).

Immunohistochemistry (IHC)
CARF expression in tissue microarray was examined as described earlier 19 . Quantitation of expression level (stained area in images) was performed using Image J software (National Institute of Health, Bethesda, MA) and represented as percentage of stained area.

Immunoblotting
Protein extraction and immunoblotting with specific antibodies (as indicated) was performed as described previously 18 . Quantitation of signals (performed with the Image J) of three independent blots (normalized with β-actin, an internal loading control) was represented as percentage expression.

Immunofluorescence (IF)
Immunofluorescence staining was performed as described earlier 18 using Zeiss Axioplan 2 microscope (with a Zeiss AxioCam HRc camera) and confocal laser scanning microscope (LSM510, Carl Zeiss). To resolve the subcellular localization of protein, the images were analyzed with the Imaris (Bitplane) software.

Invasion and wound-healing assay
Invasion assay was performed by seeding 5 × 10 4 cells in upper chamber of Matrigel (BD BioSciences) coated surface in a 12-well plate. The procedure, staining and image acquisition was performed as earlier described 19 . Wound-healing assay with different (CARF-enriched/silenced U2OS, SKOV-3, MBA-MB-231 and HT-1080) cells for 48 and 72 h was performed as described earlier 18 . Phase contrast images (×10 magnification) at different time-points were captured and quantitated for scratched area.

In vivo xenograft assays
Five-to six-weeks-old female nude mice were injected subcutaneously and intravenously with 2.5 × 10 6 control (2 groups-3 mice each; post-2 weeks for subsequent treatments, mice were randomized into 2 groups for subsequent Control and CARF siRNA treatments) and CARF-enriched (1 group; 3 mice) SKOV-3 cells. CARFtargeting adeno-oncolytic virus (RdB-shCARF) was generated 19 and employed with lung metastasis model (A549 cells, expressing firefly luciferase) as described earlier 33 . Detailed regimes of treatment, doses and tumor growth monitoring are provided in supplementary methods. All observations were performed, updated and duly discussed to avoid blinding on allocation/accessing the experimental outcome. This study was carried out in strict accordance with the recommendations in the Animal Experiment Committee, Safety and Environment Management Division, National Institute of Advanced Industrial Science & Technology (AIST), Japan (Experimental plan approval #2012-025).

Statistical analysis
All quantitative experiments were performed in triplicate. Data values are represented as mean ± SEM of three individual replicates. The two-tailed Student's t-test or Mann-Whitney U-test (non-parametric), whichever applied, was enrolled to calculate the degree of significance among the control and test group. The statistical significance was expressed as p-value ≤ 0.05; and represented by *<0.05, **<0.01, ***<0.001, while NS marked the insignificant correlation 34 .