Identification of the atypical cadherin FAT1 as a novel glypican-3 interacting protein in liver cancer cells

Glypican-3 (GPC3) is a cell surface heparan sulfate proteoglycan that is being evaluated as an emerging therapeutic target in hepatocellular carcinoma (HCC). GPC3 has been shown to interact with several extracellular signaling molecules, including Wnt, HGF, and Hedgehog. Here, we reported a cell surface transmembrane protein (FAT1) as a new GPC3 interacting protein. The GPC3 binding region on FAT1 was initially mapped to the C-terminal region (Q14517, residues 3662-4181), which covered a putative receptor tyrosine phosphatase (RTP)-like domain, a Laminin G-like domain, and five EGF-like domains. Fine mapping by ELISA and flow cytometry showed that the last four EGF-like domains (residues 4013-4181) contained a specific GPC3 binding site, whereas the RTP domain (residues 3662-3788) and the downstream Laminin G-2nd EGF-like region (residues 3829-4050) had non-specific GPC3 binding. In support of their interaction, GPC3 and FAT1 behaved concomitantly or at a similar pattern, e.g. having elevated expression in HCC cells, being up-regulated under hypoxia conditions, and being able to regulate the expression of EMT-related genes Snail, Vimentin, and E-Cadherin and promoting HCC cell migration. Taken together, our study provides the initial evidence for the novel mechanism of GPC3 and FAT1 in promoting HCC cell migration.

Liver cancer is the sixth most common cancer in terms of incidence and the fourth most common cancer-related death 1 . Primary liver cancer can be divided into hepatocellular carcinoma (HCC), cholangiocarcinoma (CCA), and mixed hepatocellular carcinoma. HCC is the most common form of primary liver cancer, accounting for about 90% 2 . Up to now, early diagnosis is the best way to control HCC. Four HCC biomarkers, alpha-fetoprotein (AFP), Golgi protein P73 (GP73), abnormal prothrombin (AP), and glypican-3 (GPC3), have been studied for early serological screening of HCC [3][4][5][6] . Further analysis is required to validate whether serum GPC3 can be used as a serological marker in HCC patients although cell surface GPC3 has been well established as a histochemical maker for HCC diagnosis 7,8 .
The Hippo/Yap pathway was initially discovered in Drosophila. The existence of this pathway was later confirmed in mammals 31 . The Hippo/Yap pathway acts through cell-cell contact inhibition 32 . The loss of contact inhibition is often seen in cancer; therefore, the Hippo/Yap pathway plays an important role in carcinogenesis. It has been established that the mammalian Hippo pathway is a potent regulator of liver organ growth, and that its www.nature.com/scientificreports/ GAT GTC TCTAC-3′ For lentivirus packaging, the packing plasmids p-Mission-Gag/Pol and p-Mission-VSV-G were mixed with expression plasmid pGreenPuro in 1 ml of Opti-Mem medium at a ratio of 3:1:1, then threefold excess of PEI solution was mixed with the plasmid solution. After the plasmids and PEI formed complex at room temperature for 20 min, the plasmid-PEI mixture were transferred to a T-75 flask of 293 T cells, gently mixed and placed in the CO 2 incubator for 3 days. The lentivirus-containing cell culture medium was collected and used to transduce target cells. Transduced cells were selected with 2 μg/ml of puromycin for 7 days to get stable knockdown cells.
Cell migration assay. Cell migration was assessed by using Transwell plate inserts with 8 µm pores (Cat.3422, Costar) following the manufacturer's instruction. Cells were harvested, counted, and added to the upper insert (2 × 10 4 /well). Fresh cell culture medium containing 20% FBS was added to the lower chamber. After 48 h, the upper chamber was washed and the un-migrated cells were removed. Cells were fixed with formaldehyde and stained with crystal violet. The staining was recorded by taking photographs and cell migration was calculated by comparison of the migrated cells between the control and treatment group.

Co-IP.
For the determination of the interaction between the endogenous GPC3 and FAT1, GPC3 from HepG2 cell lysate was pooled down with hYP7, and detection of endogenous FAT1 was visualized by HRP-conjugated anti-FAT1 antibody. Pooled human IgG (hIgG) was used as isotype control of hYP7.
To map the GPC3-binding region on FAT1, full-length GPC3 and FLAG-tagged FAT1 truncation fragments were co-expressed in 293 T cells. Cell lysate was immunoprecipitated by incubating 0.8 mg of total protein in RIPA buffer with 5 µg of the anti-FLAG antibody at 4 °C overnight. The mixture of the cell lysate and antibody was incubated with protein A/G Agarose (36403ES03, Yeasen) at room temperature for 1 h with gentle rotation. After 5 times of washing, the immune complex was recovered by boiling in SDS-PAGE loading buffer. FAT1 fragment-associated GPC3 was probed and visualized by western blot.

ELISA.
To measure the binding of GPC3 and various FAT1 fragments, recombinant His-tagged GPC3 was immobilized on a 96-well plate in PBS buffer at a final concentration of 5 μg/mL, and then blocked with 5% (w/v) BSA in PBS buffer. Various amounts of recombinant hFc-tagged FAT1 fragments was added to the plate, followed by incubation at 37 °C for 1 h to allow the binding to occur. Pooled hIgG (Cat.I4506, Sigma) was used as isotype control. After washing the plate twice with PBS buffer containing 0.05% Tween 20, the binding was detected by an HRP-conjugated goat-anti-human antibody (Cat.109-036-170, Jackson ImmunoResearch). The A 450 values were associated with the corresponding FAT1 fragment concentration, and the EC 50 values were determined by GraphPad Prism 5.0 software. Flow cytometry. To determine whether FAT1 was able to bind cell surface GPC3, flow cytometry method was used. GPC3-negative A431 and GPC3 over-expressing A431 (GPC3 + ) cells were used as tested cell lines. Cells were harvested, washed twice with PBS buffer, and resuspended in ice-cold PBS buffer containing 5% (w/v) BSA. One million of cells were incubated with 50 μg of purified FAT1-hFc fragment and hIgG isotype control (Cat.I4506, Sigma). Cell binding of FAT1-hFc fragment was detected by goat anti-human IgG conjugated with Alexa Fluor488 (ab150077, Abcam). Statistical analysis. Data analyses were performed by using the GraphPad software and expressed as the mean ± SEM. Comparisons of two groups were performed using Paired Student's t test (two-tailed). Comparisons among three or more groups were performed using one-way ANOVA. P value less than 0.05 was considered statistically significant.

Results
The interaction of GPC3 and FAT1. Given that GPC3 could modulate the Yap signaling, and FAT1 was suggested as the cell surface receptor of Yap signaling in mammalian cells, we postulated that GPC3 might interact with FAT1.
FAT1 is an enormous protein that contains 34 cadherin domains (Fig. 1A). The region (residues 3662-3788) downstream of the last cadherin was not annotated, and its crystal structure has not been determined. To visualize the possible structure and function of this region, we modeled its 3D structure by using the web tool SWISS-MODEL (https ://swiss model .expas y.org/), which predicted that the overall structure resembled the extracellular domain of human receptor tyrosine phosphatase IA-2 (insulinoma-associated protein 2) (PDB ID# 2QT7), herein we tentatively named this region RTP domain (Fig. 1B). However, it is unclear whether this RTP domain has any phosphatase activity.
To testify our prediction about the interaction of GPC3 and FAT1, we conducted co-IP to analyze the interaction between the endogenous GPC3 and FAT1. As shown in Fig. 1C, endogenous FAT1 in HCC cell lysate could be pulled down by the hYP7 antibody specific for the C-terminal epitope of GPC3.

Mapping of GPC3 binding region on FAT1.
To map the GPC3 binding region on FAT1, we constructed a series of N-terminal truncated fragments fused with FLAG-tag ( Fig. 2A). The FLAG-tagged FAT1 fragments were co-expressed with the full-length GPC3 in 293 T cells. We conducted co-IP assay to examine their interactions. As shown in Fig. 2B, we found that fragments Cad2C (covering the last two cadherin domains and the downstream whole region) and E5C (covering 1st EGF-like domain and the downstream whole region) retained the ability to co-IP GPC3, while the transmembrane-intracellular region (TMICD) and the intracellular domain www.nature.com/scientificreports/ (ICD) did not pull down GPC3, indicating the extracellular region proximal to the C-terminus of FAT1 has GPC3 binding domain.
To further narrow down the GPC3 binding region, we expressed and purified smaller FAT1 fragments as hFc fusions (named FAT1A to D) such that each fragment contains the different type of FAT1 functional or structural domains ( Fig. 2A). Fragment FAT1A covered the last four EGF-like domains, FAT1B covered the Laminin G and the 2nd EGF-like domain, FAT1C corresponded to the 1st EGF-like domain, and FAT1D corresponded to the putative RTP domain. Protein binding ELISA showed that FAT1B and FAT1D had strong GPC3 binding (Fig. 2C). FAT1A had relative weaker GPC3 binding. FAT1C had no GPC3 binding. The bind data indicates that the RTP, Laminin G, and the last four EGF-like domain may contain the GPC3 binding region.
To check the GPC3 binding specificity, flow cytometry was performed to measure the specific binding of the above-mentioned domains to GPC3 that was artificially expressed on the cell surface of GPC3-negative A431 cells. The ratio of A431 (GPC3 + ) binding versus A431 was calculated based on the geometrical mean of fluorescence intensity and was used to indicate the binding specificity of each domain (Fig. 2D,E). The results showed that FAT1A had the highest binding specificity to GPC3, with very low levels of non-specific binding to A431 cells. However, FAT1B and FAT1D had strong but non-specific binding to A431 cells to some degree, especially FAT1D. FAT1C showed neglectable binding on either A431 (GPC3+) cells or A431 cells, which is consistent with protein binding in Fig. 2C. Taken together, it clearly showed that FAT1A (the last four EGF-like domains) had the specific GPC3 binding region as summarized in Fig. 2F. Elevated expression of GPC3 and FAT1 in HCC cell lines. It was known both GPC3 and FAT is a proto-oncogene or tumorigenic gene in HCC, with low or neglectable expression in normal adult liver 3,53 . We www.nature.com/scientificreports/ compared the expression pattern of GPC3 and FAT1 in normal adult liver tissues and HCC cell lines by quantitative real-time PCR and Western blot (Fig. 3). The expression of GPC3 and FAT1 in two normal adult liver tissues could barely be detected at both mRNA (Fig. 3A,B) and protein level (Fig. 3C,D), but appeared very high on HCC cell lines. We tried to correlate the protein level of GPC3 and FAT1 in three HCC cell lines, and it seemed that GPC3 and FAT1 had a co-expression trend in HCC cells, with a R 2 value of 0.32 (Fig. 3E), implying the correlation of the two proteins. www.nature.com/scientificreports/ the expression pattern of GPC3 and FAT1 under hypoxia conditions, we treated HepG2 cells with chemical hypoxia inducer 2,2-dipyridyl (DP) as previously described 53 . As shown in Fig. 4, DP treatment significantly upregulated the expression of HIF1α, GPC3, and FAT1 at both mRNA ( Fig. 4A-C) and protein level (Fig. 4D-G). HIF1α protein level responded quickly to hypoxia induction, with an abrupt rising in 2 h of treatment. The upregulation of GPC3 and FAT1 by DP treatment was apparently lagging behind that of HIF1α.

Involvement of GPC3 and FAT1 in HCC cell migration.
Metastasis is a major attribute of cancer aggressiveness. In order to clarify the roles of GPC3 and FAT1 in HCC cell migration, we knocked down GPC3 and FAT1 expression by two shRNAs in Hep3B cells, as confirmed by Western blot (Fig. 5A,B). The effect of GPC3 and FAT1 knockdown on cell migration was determined by Transwell assay. Compared with shCtrl, both GPC3 and FAT1 knockdown significantly inhibited cell migration (Fig. 5C,D). GPC3 and FAT1 single knock down had no significant difference in suppressing cell migration, while double knock down had additive effect (Fig. 5E), indicating that the direct association of GPC3 and FAT1 might form a functional complex in suppressing migration.

PC3 and FAT1 regulated the expression of EMT related genes.
To mechanistically understand the interaction of GPC3 and FAT1 in promoting HCC cell migration, we analyzed the regulation pattern of GPC3 and FAT1 on the expression of tumor metastasis-related genes under hypoxic conditions. As shown in Fig. 6,

Discussion
Previous studies demonstrated that GPC3 promoted HCC cell migration by recruiting extracellular Wnt and HGF factors and transferring them to the corresponding receptors 23,58 . We have also demonstrated that GPC3 is involved in Yap signaling 19,22 . Since GPC3 does not have intracellular signaling domain, we postulated that GPC3 may also directly interact with some receptor-like transmembrane protein to transmit the signal. In the present study, we identified the atypical cadherin FAT1 as a new GPC3-interacting protein.
To map the GPC3 binding region on FAT1, we first focused on the functional domains, the cadherin domains, the RTP-like domain, the first EGF-like domain, the Laminin G domain, and the last four EGF-like domains, all of which are located toward the C-terminus of FAT1 extracellular region. For this purpose, the N-terminal shortened fragments that retained all or some of the functional domains were co-expressed with GPC3 in 293 T cells, and the subsequent co-IP assays determined that the cadherin repeats and the RTP-like domain may not be involved in GPC3 binding (Fig. 2B). To further identify the GPC3 binding domain on FAT1, four recombinant FAT1 fragments were expressed and purified ( Fig. 2A). The protein binding ELISA assay showed that FAT1D (the RTP domain), FAT1B (the Laminin G and the first EGF-like domain), and the FAT1A region (the last four EGF-like domains) were able to bind GPC3 (Fig. 2C), FAT1C (the first EGF-like domain alone) had no GPC3 binding. To evaluate the binding specificity on cells, we over-expressed GPC3 on A431 cells. The specific binding of FAT1 domains on A431 (GPC3+) versus A431 was compared. It showed that FAT1A had the most specific binding to A431 (GPC3), while FAT1B had less specific binding to GPC3 since it also bound to A431 cells. The FAT1C did not bind to the cell, which is consistent with our ELISA data. The FAT1D had no specificity for GPC3 binding because it bound to A431 (GPC3) equally strong as A431 (Fig. 2D,E). Taken together, it is clear that the last four EGF-like domains represented by FAT1A (residues 4013-4181) on FAT1 have the specific GPC3 binding domain (Fig. 2F). Laminin G and the RTP domain could bind GPC3 on cells nonspecifically. Interestingly, the first EGF-like domain as represented in FAT1C did not bind GPC3.
In view of previous studies, both GPC3 and FAT1 are highly expressed in HCC and correlated with poor prognosis, and promote HCC cell migration and proliferation 24,53,59,60 . Here, we found that high-level expression www.nature.com/scientificreports/ of GPC3 coincided with that of FAT1 in all the tested liver cancer cell lines (HepG2, Hep3B, and Huh7), and undetectable in normal liver tissues (Fig. 3). Moreover, impaired GPC3 and FAT1 expression by shRNA knockdown suppressed HCC cell migration in a comparable level (Fig. 5), although double knockdown of GPC3 and FAT1 had a slightly greater inhibition of HCC cell migration compared to single knockdown (Fig. 5). Taken these data with previous findings that both GPC3 and FAT1 were able to modulate the activity of YAP, the downstream effector of Hippo pathway 35,61 , it would be reasonable to postulate that GPC3 and FAT1 might function as complex in promoting HCC cell migration. Hypoxia is a major feature of tumor microenvironment and a driving force to promote cancer metastasis, including HCC 55,[62][63][64] . Hypoxia induces the expression of a number of genes [65][66][67] , and some of them drive tumor metastasis, e.g. HIF1α 68 . FAT1 is another important gene that regulates EMT and stemness characteristics in hypoxia GBM tumor 43 and drives tumor metastasis 53 . The current study used DP treatment to simulate hypoxia environment as previously described 53 , and it was found that DP treatment significantly up-regulated the expression of HIF1α, GPC3, FAT1, and other tumor metastasis-related genes (Snail, Vimentin) (Figs. 4,6). Furthermore, both GPC3 and FAT1 regulated the expression of tumor metastasis-related genes, e.g., Snail, Vimentin, and E-Cadherin (Fig. 6D-F). As expected, single knockdown of either GPC3 or FAT1 suppressed the regulation of metastasis genes Snail, Vimentin, and E-Cadherin, and double knockdown of GPC3 and FAT1 had a little enhanced effect on the regulation of some of the EMT genes (Fig. 6), suggesting again that GPC3 and FAT1 may work via a shared mechanism to promote HCC cell migration.
In conclusion, the current work identified the direction interaction of transmembrane protein FAT1 and GPC3. In support of this observation, FAT1 and GPC3 had similar expression patterns and functional features in terms of promoting HCC cell migration, and both GPC3 and FAT1 could regulate the expression of metastasisrelated genes.

Data availability
Data sharing is not applicable to this article as no datasets were generated or analyzed during the current study. Materials from the present study are available from the corresponding author on reasonable request. www.nature.com/scientificreports/