An Oncogenic Protein Golgi Phosphoprotein 3 Up-regulates Cell Migration via Sialylation*

Background: Molecular mechanisms of the effect of the GOLPH3 oncogenic protein on tumorigenesis remain unclear. Results: GOLPH3 specifically up-regulates sialylation of integrin N-glycans, promotes sialylation-dependent cell migration, and affects AKT signaling. Conclusion: GOLPH3 affects cell biological functions through a specific regulation of sialylation. Significance: The sialylation of N-glycans is important for functions of GOLPH3. Recently, the Golgi phosphoprotein 3 (GOLPH3) and its yeast homolog Vps74p have been characterized as essential for the Golgi localization of glycosyltransferase in yeast. GOLPH3 has been identified as a new oncogene that is commonly amplified in human cancers to modulate mammalian target of rapamycin signaling. However, the molecular mechanisms of the carcinogenic signaling pathway remain largely unclear. To investigate whether the expression of GOLPH3 was involved in the glycosylation processes in mammalian cells, and whether it affected cell behavior, we performed a loss-of-function study. Cell migration was suppressed in GOLPH3 knockdown (KD) cells, and the suppression was restored by a re-introduction of the GOLPH3 gene. HPLC and LC/MS analysis showed that the sialylation of N-glycans was specifically decreased in KD cells. The specific interaction between sialyltransferases and GOLPH3 was important for the sialylation. Furthermore, overexpression of α2,6-sialyltransferase-I rescued cell migration and cellular signaling, both of which were blocked in GOLPH3 knockdown cells. These results are the first direct demonstration of the role of GOLPH3 in N-glycosylation to regulate cell biological functions.

Protein glycosylation is one of the most prevalent forms of post-translational modification, and altered glycosylation is a hallmark feature of cancers (1). Integrin is one of the major carriers of N-glycans in its governing of cell migration, proliferation, and differentiation (2). Integrin-mediated biological functions such as cell spreading and cell migration can be modulated as a consequence of an aberrant change in the N-glycosylation of integrins, which is often associated with the meta-static process. A series of studies (including by our group) have reported that alterations in the oligosaccharide portion of integrins that are modulated by the expression of each glycosyltransferase gene, such as N-acetylglucosaminyltransferases III (GnT 2 -III, also called MGAT3) and V (GnT-V, also called MGAT5), as well as by sialyltransferases, regulate cell malignant phenotypes such as integrin-mediated cell migration and cell spreading. For example, the expressions of GnT-V and ␤1,6-branched N-glycan levels are increased in highly metastatic tumor cell lines (3,4), which enhances integrin-mediated cell migration. In contrast to GnT-V, the overexpression of GnT-III blocks branched N-glycan and results in an inhibition of integrin-mediated cell spreading and migration as well as in the phosphorylation of the focal adhesion kinase, thereby contributing to the suppression of cancer metastasis (5,6).
Besides the branched N-glycans, the terminal sialic acids are believed to be common cancer-associated carbohydrate modifications (7). Sialic acids also are believed to be essential for the early development of vertebrates (8). Enhanced expression of ␣2,6-linked sialylation on N-glycans often correlates with human cancer progression, metastatic spread, a poor prognosis, and stem cell markers (7,9,10). Increased expression has been reported in carcinomas of the colon (11), breast (12), cervix (13), and ovary (9) and in some brain tumors (14). The expression level of the ␣2,6-sialyltransferase-I (ST6GAL1) gene is up-regulated by the Ras oncogene (15,16). However, the molecular mechanism for the post-translational regulation of sialylations in cancer cells remains unclear.
The increased ␣2,6-linked sialylation on ␤1-integrins also has been reported in several transformed cell types and is postulated to alter integrin function by enhancing its activation state and binding to collagen (17)(18)(19). In these studies, increases in ␣2,6-linked sialylation levels have been correlated with enhanced cell motility and invasiveness in vitro. Furthermore, the role of ST6GAL1 enzyme has been confirmed in in vivo growth and differentiation, for which ␤1-integrin function is important for tumorigenesis and in maintaining the proliferative state of tumor cells (20). Thus, the state of the N-glycans of integrin plays important roles in a poor prognosis for cancer, such as cell-cell adhesion, cell-extracellular matrix interaction, epithelial-mesenchymal transition, and metastatic ability.
Most of the glycosylation reactions happen in the Golgi apparatus. Recently, a Golgi phosphoprotein 3 (GOLPH3) was identified as an oncogenic protein in human solid tumors such as lung cancer, breast cancer, colon cancer, and melanoma, localized on the peripheral membrane of the trans-Golgi network and modulating a mammalian target of rapamycin (mTOR) signaling (21), budding of vesicles from the trans-Golgi, and recycling of transmembrane receptors (22). It is worth noting that the expression levels of GOLPH3 are highly related to the clinical stages of breast (23), esophageal (24), and lung (21) cancers and glioblastoma (25). The homolog of yeast GOLPH3, VSP74, is reportedly involved in the retention of mannosyltransferases at the Golgi, knockout of which results in the production of hypoglycosylated proteins (26). Recently, Ali et al. (27) reported that GOLPH3 regulates the Golgi retention of the O-glycan synthesis enzyme, core2 glucosaminyl(N-acetyl)transferase 1 (GCNT1), in mammalian cells because it has a similar sequence to the yeast's mannosyltransferases that are present in the cytoplasmic tail. However, the effects of GOLPH3 on N-glycanosylation remain mostly unclear.
In this study, we performed a knockdown and restoration of the GOLPH3 gene in mammalian cells to investigate the effects of GOLPH3 on N-glycanosylation and its related biological functions. We found that GOLPH3 specially regulates sialylation of N-glycans and integrin-mediated cell migration, which may provide new insight into the functions of GOLPH3 in cancer.
Cell Lines and Cell Culture-The HeLa and 293T cells were provided from RIKEN Cell Bank (Japan). The Phoenix cells and MDA-MB231 cells were purchased from ATCC. All cell lines were maintained at 37°C in Dulbecco's modified Eagle's medium (DMEM), supplemented with 10% fetal bovine serum (FBS), 100 units/ml penicillin G, and 0.1 mg/ml streptomycin, under a humidified atmosphere containing 5% CO 2 , except for the virus production.
Virus Infections-Viral infection was performed as described previously (33,34). In brief, the lentivirus vectors (CSIV-TRE-RfA-CMV-KT or CS-RfA-ETBsd) were transfected into 293T cells with packaging plasmids by calcium phosphate. The pBABE hygro-RfA or pSUPER.retro.puro were transfected into Phoenix cells for the retrovirus. The target cells were cultured for 24 h to obtain virus media for infection. After infection for 72 h, cells were selected by the FACSAria II (BD Biosciences) to obtain Kusabira Orange-positive cells (CSIV-TRE-RfA-CMV-KT) or selected by puromycin (pSUPER.retro.puro), blasticidin (CS-RfA-ETBsd), or hygromycin (pBABE-hygro-RfA) to get resistant cells against these antibiotics.
Cell Migration-Each Transwell (BD BioCoat TM control inserts, 8.0-m inserts; BD Biosciences) was coated only on the bottom side with 10 g/ml FN at 37°C for 1 h. Cells were trypsinized, and the trypsin was neutralized with 1 g/ml soybean trypsin inhibitor, and cells were resuspended in DMEM. The suspended cells were centrifuged, and the cell pellets were resuspended in an assay medium (0.1% BSA in DMEM containing 3% FBS) and diluted to 2 ϫ 10 6 cells/ml; cell viabilities were confirmed by trypan blue staining. Cell suspensions of 500-l aliquots were added to each FN-coated transwell, followed by incubation at 37°C for 6 h for HeLa cells and 3 h for MDA-MB231 cells. After incubation, cells on the upper side were removed by scraping with a cotton swab. The membranes in the transwells were fixed with 4% paraformaldehyde and stained with 0.5% crystal violet for 30 min. Cells that had migrated to the lower side were counted using a phase-contrast microscope.
Video Microscope-A glass-bottom dish (Asahi Techno Glass, Japan) was precoated with 1 g/ml LN-332 (36) (a precious gift from Dr. K. Miyazaki, Yokohama City University, Japan) and then blocked with 1% BSA for 1 h at 37°C. A 200-l aliquot of the cell suspension (4 ϫ 10 4 cells/ml) in growth medium was added to each LN-332-precoated glass-bottom dish. Cell migration was monitored for 8 h using time-lapse video equipment (Carl Zeiss, Germany). After incubation for 1 h, images and movies were acquired using inverted microscopes (Axio Observer.D1, Carl Zeiss) every 5 min at 37°C with 5% CO 2 in a heated chamber with temperature and CO 2 controller (Onpu-4 & CO2, Air Brown, Japan) during time-lapse imaging. Cell migration was evaluated using Axio Vision Rel.

(Carl Zeiss).
Preparation of Pyridylaminated N-Linked Oligosaccharide and Analysis of N-Glycans by the Reversed-phase HPLC-N-Glycan analysis was performed with minor modification as reported previously (37). The cells (2-4 ϫ 10 7 cells) were lysed by homogenization, and the N-glycans were then released with N-glycosidase F (Roche Applied Science) from 10 mg of proteins of cell lysates by incubation for 16 h. The pyridylaminated N-glycans (PA-N-glycans) were prepared according to the manufacturer's recommended procedure (pyridylamination manual kit, Takara Bio Inc., Shiga, Japan). Excess 2-aminopyridine was removed using a cellulose cartridge. The PA-N-glycans prepared from the cells were analyzed using a reversedphase HPLC system (Shimazu Co., Kyoto, Japan) with an ODS80-TM column (4.6 ϫ 150 mm; Tosoh Corp., Tokyo, Japan). Elution was performed at a flow rate of 1.0 ml/min at 55°C using 20 mM ammonium acetate buffer, pH 4.0, as solvent A and the same buffer containing 1% 1-butanol as solvent B. The column was pre-equilibrated with 4% solvent B, and after injection of a sample, the PA-N-glycans were separated using  Fig. 1C pSUPER.retro.puro b Fig. 2 pSUPER.retro.puro b Fig. 3, A-C pSUPER.retro.puro b pBABE hygro-RfA Fig 4% of solvent B for 10 min and then a linear gradient of 4 -30% of solvent B for 60 min. PA-N-glycans were detected using a fluorescence detector at excitation and emission wavelengths of 320 and 400 nm, respectively. The elution time for each pyridylaminated N-glycan of HPLC was standardized using pyridylaminated (NeuNAc-Gal-GlcNAc) 2 Man 3 (GlcNAc) 2 and pyridylaminated (Gal-GlcNAc) 2 Man 3 (GlcNAc) 2 N-glycans (Takara Bio). N-Glycan Profiling by LC/MS-N-Glycan profiling was performed with minor modification as reported previously (38). Cell pellets (1 ϫ 10 7 cells) were homogenized in 2 ml of TBS (20 mM Tris-HCl, pH 7.4, 150 mM NaCl) containing protease inhibitor mixture (Nacalai Tesque) using a Potter-Elvehjem homogenizer. The homogenized cell lysates were centrifuged at 1000 ϫ g for 10 min at 4°C, and the membrane fractions were obtained by an ultracentrifugation of the supernatants at 120,000 ϫ g for 80 min at 4°C. N-Glycosidase F digestion was performed overnight on the membrane fraction at 37°C, according to the manufacturer's recommended procedure. The released N-glycans were reduced in 500 l of 0.5 M NaBH 4 at room temperature for 16 h, and neutralized using 5% acetic acid. The reduced N-linked glycans were recovered using a solid-phase extraction cartridge (EnviCarb C, Supelco, Bellefonte, PA) and were lyophilized. The reduced N-linked glycans were separated on a graphitized carbon column (Hypercarb, 150 ϫ 0.1 mm, 5 m; Thermo Fisher Scientific, Waltham, MA) at a flow rate of 500 nl/min in an UltiMate 3000 RSLCnano LC system (Dionex, Sunnyvale, CA). The mobile phases were 5 mM ammonium bicarbonate containing 2% acetonitrile (A buffer) and 5 mM ammonium bicarbonate containing 80% acetonitrile (B buffer). The glycans were eluted with a linear gradient of 5-55% of B buffer for 60 min. Mass spectrometric analysis was performed using a Fourier transform ion cyclotron resonance/ ion trap-type mass spectrometer (FT-MS, LTQ-FT; Thermo Fisher Scientific). For mass spectrometry, the electrospray voltage was 2.5 kV in both the positive and negative ion modes. The resolution of FT-MS was 50,000, and the scan range was m/z 700 -2000. The monosaccharide compositions of the glycans were deduced based on an accurate measurement of the mass, as obtained by FT-MS.
Flow Cytometric Analysis-Flow cytometric analysis was performed as described previously with minor modifications (5,33,35). Briefly, semi-confluent cells were detached from the culture dishes using trypsin containing 1 mM EDTA and were sub-sequently stained with or without the primary mouse anti-␤1, followed by incubation with Alexa Fluor 647 goat anti-mouse IgG (Invitrogen). For lectin staining, cells were stained by biotinylated MAM or Ricinus communis agglutinin I (RCA-I), followed by incubation with Alexa Fluor 647-conjugated streptavidin. After washing three times with PBS, flow cytometric analysis was performed using a FACSCalibur flow cytometer and Cell Quest Pro software (BD Biosciences).
Xenograft Tumor Formation-To evaluate tumor formation in the wild-type and GOLPH3-KD HeLa cells, the athymic nude mice were given a subcutaneous injection of those cells. Briefly, the cells were plated on a 100-mm dish in complete medium. After treatment with PBS containing 1 mM EDTA, the cells were suspended to a single-cell level with DMEM. A total of 1 ϫ 10 6 cells were subcutaneously injected into the NOD-athymic mice (5-week-old male BALB/c mice; Charles River Laboratories, Japan). 6 weeks after the injection, the tumor sizes were measured (n ϭ 5 per group). Tumor volume was calculated using the formula V ϭ (L ϫ W 2 ) ϫ 0.5 ( where V ϭ volume, L ϭ length, W ϭ width). The tumor weight was also measured after harvesting. All animal procedures were carried out according to experimental protocols approved by the Tohoku Pharmaceutical University Research Ethics Board.
Statistical Analysis-Statistical analyses were performed using either a Student's t test or one-way analysis of variance (ANOVA), using GraphPad Prism5.
Confocal Microscope-The expression vector of pEF puro ST6GAL1-GFP was transfected into the doxycycline-inducible GOLPH3 knockdown cells. After selection with 1 g/ml puromycin, cells were grown on the coverslips (MatTek Corp., Ashland, MA) in the presence (KD) or absence of doxycycline for 72 h. The coverslips were rinsed twice with PBS and fixed with 4% paraformaldehyde for 10 min. The cells were then incubated with 0.1% Triton X-100 for 5 min at room temperature. Nonspecific interactions were blocked with 5% BSA at 37°C for 60 min. After rinsing three times with PBS, cells were incubated with mouse anti-GM130, a Golgi marker, mAb (BD Biosciences, 610823) and then with secondary Alexa 568-labeled goat anti-mouse IgG and TO-PRO-3 (Molecular Probes) in the dark. Samples were analyzed by confocal microscopy using an FV-1000 confocal microscope (Olympus, Tokyo, Japan).

Expression of GOLPH3 Is Important for Integrin-meditated
Cell Migration and Tumor Formation-GOLPH3 modulates the phosphorylation status of the AKT-mTOR signal pathway (21), which is a central regulator of cell growth, proliferation, differentiation, and survival. Recent studies have shown that mTOR also plays a critical role in the regulation of tumor cell motility, invasion, and cancer metastasis (41). However, the role of GOLPH3 in cancer cell migration and metastasis remains largely unknown.
To understand the effects of GOLPH3 expression on cell migration, we made knockdown HeLa cells by transfection with the siRNA-targeting GOLPH3 gene, and examined cell migration using a Boyden chamber trans-migration assay. The knockdown efficiency of GOLPH3 was verified via immunoblotting analysis (Fig. 1A). As shown in Fig. 1B, the depletion of

Effect of GOLPH3 on N-Glycosylation
GOLPH3 dramatically inhibited cell migration on FN. The cell migration was completely suppressed by the addition of anti-␤1 integrin-blocking antibody, suggesting that the expression of GOLPH3 plays an important role in integrin-mediated cell migration. In contrast to cell migration, there were no significant differences in cell viabilities between control and KD cells confirmed by trypan blue staining (data not shown). To determine the effects of GOLPH3 on tumor growth, wildtype or GOLPH3-KD cells were injected into 5-week-old male athymic mice, and tumor growth was monitored. As shown in Fig. 1C, control cells permitted the vigorous formation of tumors. However, the tumor formation was significantly suppressed in the KD cells, which was consistent with previous reports (21,23). These results suggest that GOLPH3 involved tumor formation and malignant transformation both in vitro and in vivo.
Sialylation Was Decreased in the GOLPH3-Knockdown Cells-Several studies, including those conducted by our group, have reported that the N-glycans of integrins affect cell migration. For example, ST6GAL1 knock-out mice experienced an enhanced epithelial tumor differentiation through the reduction of ␤1 integrin-mediated signaling (20). To further examine the effects of GOLPH3 on ␤1 integrin-mediated functions and N-glycosylation, we established stable GOLPH3 knockdown cell lines ( Fig. 2A, lower panels). It is noteworthy that the band mobility of ␤1 integrin on SDS-PAGE differed between the control and the knockdown cells. To clarify whether the difference of ␤1 band mobility on SDS-PAGE was due to N-glycosylation or core protein degradation, the immunoprecipitated ␤1 integrins were treated with peptide:N-glycosidase F for deglycosylation. The treatment completely diminished the differ-ence in the band mobility of ␤1 on SDS-PAGE between both cells, indicating that the knockdown of GOLPH3 affects the N-glycosylation of ␤1 integrin ( Fig. 2A, upper panels). To further explore the difference in N-glycosylation, we analyzed pyridylaminated N-glycans that had been obtained from control and knockdown cells by using reversed-phase HPLC. It is interesting that the peaks eluting at around 50 min were dramatically decreased in the GOLPH3 knockdown cells, compared with those in control cells (Fig. 2B, upper panel). These peaks disappeared after digestion with sialidase (Fig. 2B, middle panels). After sequential treatments with sialidase and ␤-galactosidase, the elution patterns of the labeled N-glycans could not be distinguished between the GOLPH3 knockdown cells and the control cells (Fig. 2B, middle and lower panels). These results strongly indicated that GOLPH3 specifically affected sialylation on the N-glycans.
The sialylation of N-glycans is ordinarily accomplished using ␤-galactoside ␣-2,3-sialyltransferase-III, IV (ST3GAL3 and ST3GAL4), and ␤-galactoside ␣-2,6-sialyltransferase-I (ST6GAL1). In contrast, neuraminidase1 (NEU1) reduces the sialylation of N-glycan on the cell surface. We used RT-PCR to confirm whether the expression of GOLPH3 affected the gene expression levels of these sialyltransferases and sialidase. As shown in Fig. 2C, there were no significant differences between the control and the GOLPH3 knockdown cells, suggesting that the sialylation was regulated by GOLPH3 rather than by the gene expression of the sialytransferases and the sialidase.
Restoration of GOLPH3 Expression Rescued the Sialylation and Cell Migration, Which Were Blocked in the GOLPH3 Knockdown Cells-Many studies have suggested that sialylation played a crucial role in integrin-mediated cell migration  (18,20,42). To clarify the relationships between GOLPH3 and sialylation, as described above, we overexpressed an shRNAresistant GOLPH3 gene in the GOLPH3-knockdown cells and then examined cell migration and N-glycan structures. As shown in Fig. 3A, the overexpression of GOLPH3 in the knockdown cells greatly rescued integrin-mediated cell migration on LN-332, which was blocked in the GOLPH3 knockdown cells. As expected, the mobility on SDS-PAGE for the ␤1 integrin band obtained from the rescue cells was slower than that from the KD cells and returned to a state similar to that of the control cells (Fig. 3B). It is noteworthy that the treatment with sialidase cancelled these differences.

Effect of GOLPH3 on N-Glycosylation
To further confirm the difference, the N-glycan profiles of those three cells were compared using LC/MS. The data showed that sialylated N-glycans were decreased, and asialo-Nglycans were increased in KD cells, compared with those in the control cells (Fig. 3C), which was consistent with the results of HPLC analysis, as described in Fig. 2B. Furthermore, after the re-introduction of GOLPH3 (rescue cells), the expression pattern of sialylated and asialo-N-glycans in the KD cells was almost normalized to that of the control cells, suggesting that GOLPH3 specifically regulates sialylation of N-glycans. To examine directly the effects of GOLPH3 on sialylation, we designed a Tet-On expression system for knockdown and overexpression of GOLPH3. A positive correlation between the levels of the sialylated ␤1 integrin and the expression levels of GOLPH3 was also observed in the GOLPH3 knockdown and overexpression cells (Fig. 3D). However, the expression levels of GOLPH3 did not significantly affect ␤1 integrin expressed on the cell surface, which was confirmed by FACS analysis (Fig.  3E). The down-regulation of sialylation was also observed in GOLPH3 knockdown MDA-MB231 cells (Fig. 3F). Furthermore, a suppression of cell migration was also observed in GOLPH3 knockdown MDA-MB231 cells (Fig. 3G). To analyze overall sialylation, we performed FACS analysis. The downregulation of sialylation was also observed in FACS analysis (Fig. 3H). The reactivity against MAM lectin was decreased, although the reactivity against RCA-I lectin was increased in GOLPH3 KD HeLa cells, compared with control cells. These results were consistent with the data analyzed by HPLC and MS (Figs. 2B and 3C). These results, taken together, strongly suggest that GOLPH3 influences sialylation.
To examine whether GOLPH3 specifically regulates the sialylation on ␤1 integrin, we also investigated other glycoproteins. It was interesting that a clear correlation of GOLPH3 expression with enhanced sialylation was also observed in ␣5 integrin but not in either ␣3 integrin or EGFR. A decreased GOLPH3 expression significantly suppressed both ␣2,6and ␣2,3-sialylated integrin ␣5, although an increased GOLPH3 expression greatly enhanced both sialylations (Fig. 3I). The GOLPH3 expression did not affect the ␣2,3-sialylation of integrin ␣3 (Fig. 3J). Although it is possible that the overexpression of GOLPH3 could increase the sialylated EGFR, the knockdown of GOLPH3 only modestly affected the sialylation (Fig. 3K). Overexpression of GOLPH3 promoted cell migration as shown in Fig. 3L. Collecting these results suggests that the decreases in sialylation were different among different glycoproteins. Although the underlying molecular mechanism remains unclear, we could spec-  JULY 25, 2014 • VOLUME 289 • NUMBER 30 ulate that GOLPH3 differentially affects sialylation on target proteins.

Effect of GOLPH3 on N-Glycosylation
Forced Expression of ST6GAL1 Rescued Cell Migration and AKT Phosphorylation, Which Were Blocked in KD Cells-As described above, the gene knockdown of GOLPH3 suppressed cell migration and sialylation, suggesting GOLPH3 is involved with cell migration via integrin and sialylations. GOLPH3 can also regulate downstream growth signaling in response to receptor tyrosine kinase activation (21,43). Therefore, we hypothesized that GOLPH3 might affect the PI3K-AKT-

Effect of GOLPH3 on N-Glycosylation
mTOR signaling pathway through the sialylations. In fact, many studies have reported that the sialylated ␤1 integrin plays important roles in its biological functions (18,20,44). Here, we examined whether a forced expression of sialyltransferases overcomes those phenotypes observed in GOLPH3 knockdown cells. As shown in Fig. 4, the overexpression of ST6GAL1 rescued not only integrin-mediated cell migration (Fig. 4A) but also EGFR-mediated AKT phosphorylation, both of which were suppressed in GOLPH3 knockdown cells (Fig. 4C). However, it is worth noting that the overexpression of ST3GAL4 could partly rescue EGFR-mediated AKT phosphorylation (data not shown) but not efficiently rescue the deficiencies of cell migration of GOLPH3 knockdown cells (Fig. 4B). The ␣2,6-sialylation status of ␤1 integrin was greatly increased in GOLPH3-KD cell by overexpression of ST6GAL1 (Fig. 4D). The total ␣2,6sialylations levels confirmed by FACS analysis using SSA lectin were also increased (data not shown). These data suggest that the enhanced expression of sialylated N-glycans by GOLPH3 could be one of the mechanisms for its oncogenic signaling. GOLPH3 is known to regulate mTOR signaling (21). Unexpectedly, the treatment with rapamycin greatly up-regulated the expression of sialylated N-glycans on ␤1 integrin (Fig. 4E).
GOLPH3 Was Specifically Associated with Sialyltransferases-In yeast cells, Vps74p, which is a counterpart of GOLPH3, was associated with mannosyltransferase enzyme (26), coatomer (45), phosphatidylinositol 4-phosphate (PI4P) (29), and regulated N-glycan synthesis. Very recently, GCNT1 enzyme was reported to be one of the putative binding partners for GOLPH3 to retain its Golgi localization (27). To understand how GOLPH3 could regulate sialylation specifically, we tested whether GOLPH3 binds to sialyltransferase(s) such as ST3GAL4 and ST6GAL1, which mainly participate in the sialylation of N-glycans in HeLa cells. The co-immunopre-cipitation showed that GOLPH3 specifically associated with ST3GAL4. Furthermore, the chimeric ST3GAL4 containing the cytoplasmic domain of ␤4GALT1 (Fig. 5A) lost its binding, although the chimeric ␤4GALT1 became associated when its cytoplasmic domain was replaced by ST3GAL4 (Fig. 5B). The association of GOLPH3 and ST6GAL1 was also observed (Fig.  5C). These results suggest that GOLPH3 is able to bind to a specific region within the cytoplasmic domains of sialyltransferases. In fact, GOLPH3 recognized PI4P via a positively charged binding pocket on the hydrophobic face of the protein (22,29) for localization in the trans-Golgi. Vps74p oligomer is reported to be required for Golgi localization of glycosyltransferases in yeast cells (16). Therefore, we next sought to determine whether a lack of PI4P binding ability (R171A/R174A or W81A/R90A) (29) or the disruption of its oligomerization (⌬190 -201) (16) rescued the aberrant sialylation observed in GOLPH3 knockdown cells. As expected, these GOLPH3 mutants could not rescue the sialylation levels of ␤1 integrin, suggesting the GOLPH3/sialyltransferase/PI4P participated in the regulation of ␣2,6-sialylations in a coordinated fashion (Fig. 5D).
To examine the effects of GOLPH3 on localization of sialytransferases, we used a doxycycline-inducible system to establish the GOLPH3-KD cells expressed with ST6GAL1-GFP. To be consistent with the previous report (22), the localization of ST6GAL1-GFP did not apparently change in the presence or absence of doxycycline (Fig. 5E).

DISCUSSION
In this study, we found the following: (i) that GOLPH3 expression played an important role in integrin-mediated cell migration via the up-regulation of sialylation; (ii) that GOLPH3 was specifically associated with sialyltransferases and the regulated sialylation of N-glycans on ␤1 integrin; and (iii) that the expression of ST6GAL1 rescued the integrin-mediated cell FIGURE 3. Effects of restoration of GOLPH3 expression on ␤1 integrin-mediated cell migration and its N-glycosylation. A, GOLPH3 knockdown cells as described in Fig. 2A were infected with retrovirus-expressing shRNA, a resistant GOLPH3 gene, and then cultured in the presence of hygromycin to achieve 100% infection, which was used as rescued cells (Res). Cell migration on laminin 332 (upper panel) was monitored by time-lapse microscopy as described under "Experimental Procedures." Each bar represents the means Ϯ S.D. of the migration distance of 10 cells in each assay (*, p Ͻ 0.0001, by one-way ANOVA with Tukey's post hoc test). The expression levels of GOLPH3 were compared among the control (Ctrl), the KD, and the rescued cells (middle panel), which were restored with the expression of GOLPH3 in KD cells. The ␣-tubulin was used as a loading control (lower panel). B, Western blotting for ␤1 integrin. The same amounts of cell lysates (200 g) obtained from the indicated cells were immunoprecipitated (IP) with anti-␤1 antibody, and the immunoprecipitates were digested with or without sialidase. After the treatment, the immunoprecipitates were then immunoblotted with ␤1 antibody. Ctrl, control shRNA; KD, GOLPH3knockdown; Res, KD cells overexpressed a shRNA-resistant GOLPH3 gene. C, comparison of major N-glycans from different cells by glycan profiling using LC/MS peak area of asialo and sialylated N-glycans were calculated using mass spectra obtained in positive and negative ion modes, respectively. The relative peak area of major N-glycans from the control, KD, and the rescued cells were expressed as a percentage of the total peak area of the glycans. Glycan structures were deduced by MS analysis. D, proteins extracted from HeLa cells that expressed a Tet-On expression system for shRNA of GOLPH3-292 (left panel) or the GOLPH3 gene (right panel) using lentivirus systems as described under "Experimental Procedures" with or without doxycycline (DOX), were immunoprecipitated with the indicated lectin, resolved by SDS-PAGE, and then immunoblotted for ␤1 integrin antibody. Cell lysates were also Western blotted for the indicated antibodies. SSA is an ␣2,6-sialic acid-specific lectin, and MAM is an ␣2,3-specific lectin.

Effect of GOLPH3 on N-Glycosylation
migration and the intracellular signaling of cancer cells, both of which were down-regulated in the GOLPH3-KD cells. These observations are the first to directly demonstrate the role of GOLPH3 in the N-glycosylation of glycoproteins to regulate the signaling events, which may influence mTOR signaling and tumor progression (Fig. 6).
Integrins are cell-surface glycoproteins that mediate cell-ECM interactions and link matrix proteins to the cytoskeleton (2). They play an important role in intracellular signal transduction (46), regulating various processes, such as cell proliferation, differentiation, apoptosis, and cell migration. Most integrins are major carriers of N-glycans, and changes in these structures can alter the cell-cell and cell-ECM interactions, thereby affecting cell adhesion, migration, and tumor malignancy. The ␤1 integrin heterodimerizes with one of 12 possible ␣ subunits and mediates adhesion, spreading, and migration on multiple ligands, including collagen, laminin, and fibronectin (47,48). A recent study on the crystal structure of ␣5␤1 ectodomain showed that the RGD-binding pocket is surrounded by several N-glycan chains, leaving an exposed surface along the subunit interface (49). Computer modeling also showed that sialylation on an I-like domain may affect the signaling mediated by integrins (44). These results strongly support the notion obtained from this and previous studies, in which the alteration in the N-glycans of ␣5␤1 on the cell surface could affect the biological function of the receptor (33,35). In fact, the importance of individual N-glycans and the N-glycosylation sites of integrins have gradually become clearer. As described above, integrin is ideally suited to the influence of tumor cell behavior in diverse extracellular matrix milieus, and its N-glycosylation state is changed in physiological and pathological conditions such as tumorigenesis and cancer metastasis. However, the underlying molecular mechanism of these changes remains largely unexplained.
The original function of the GOLPH3 yeast homolog Vps74p as a protein that retains mannosyltransferases in the Golgi by interacting with a consensus amino acid sequence (26) had been observed in mammalian cells, in which GOLPH3 was bound to GCNT1 and regulated the cell-cell interaction through O-glycans (27). In this study, we found that GOLPH3 specifically associated with sialyltransferases and then affected the sialylation of N-glycans. Sialyltransferases, however, do not contain the consensus amino acid sequence that is shown in GCNT1, which suggests the existence of a novel molecular mechanism for the regulation by GOLPH3. Taken together, we could speculate that the direct or indirect association among GOLPH3, PI4P, and sialyltransferases efficiently regulates sialylation, especially ␣2,6-sialylation. For sialylation, an interac-tion with PI4P, which is enriched in the trans-Golgi apparatus, seems to be very important because the GOLPH3 mutant, which lacks PI4P binding ability, was unable to rescue the ␣2,6sialylation of ␤1 integrin. The interaction between GOLPH3  . Proposed molecular mechanism for the regulation of sialylation and cellular signaling by GOLPH3. GOLPH3 has been known to modulate mTOR signaling (21). This study clearly showed that the interaction among GOLPH3, PI4P, and sialyltransferases might efficiently regulate ␣2,6-sialylation on several target proteins expressed on the cell surface, including integrins and some receptor tyrosine kinases (RTK). Then those resultant glycoproteins could cooperatively enhance integrin-mediated cell migration and activate cellular signal pathways such as the phosphoinositide 3-kinase (PI3K)-AKT-mTOR cascade. The possible molecular mechanism described here might partly explain the observation that the GOLPH3 gene is usually amplified in many malignant tumors. and the cytoskeleton via unconventional myosin MYO18A and PI4P was reportedly involved in the exocytosis and maintenance of the Golgi (22). Here, we observed no significant change in the expression levels of ␤1 integrin on the cell surface with either the knockdown or the overexpression of GOLPH3 (Fig. 3E). However, we also could not detect any significant changes in the overall expression level of sialylated N-glycans between control and GOLPH3 overexpression cells (data not shown). The possible mechanisms could be explained as follows. First, the overexpression of GOLPH3 might selectively affect only a few glycoproteins, such as ␤1 integrin, and not all glycoproteins. Second, the overexpression of GOLPH3 might efficiently affect ␣2,6-sialylation, but not ␣2,3-sialylation, which is a larger portion of sialylation. In fact, it has been reported that the expression levels of ␣2,3-sialylation were much higher (ϳ6.5-fold) than the ␣2,6-sialylation on ␣5␤1 integrin obtained from placenta (50).
GOLPH3 has been known to modulate mTOR signaling and oversensitivity for rapamycin. High GOLPH3 expression correlates with the hyperactivation of mTOR signaling in human cancer cells. Unexpectedly, we found that treatment with rapamycin up-regulated the sialylation of ␤1 integrin (Fig. 4E). The details of the mechanism remain unclear. It could be speculated that cancer cells may escape from the apoptosis induction by rapamycin treatment and achieve cell survival and cell invasion through an up-regulation of sialylation. In fact, Ma et al. (51) reported that the expression ST6GAL1 was up-regulated in multidrug-resistant tumor cells. Thus, we believe that the induction of sialylation could allow tumor cells to acquire new potentials for malignancy and recurrence in some clinical cases. Furthermore, overexpression of ST6GAL1 efficiently rescued cell migration and AKT phosphorylation, which were blocked in the GOLPH3-knockdown cells (Fig. 6).
The results of this study clearly showed GOLPH3-mediated sialylation induction and the attendant biological functions. Specifically, ␤1 integrin was one of the important targets for sialylation. The knockdown of GOLPH3 resulted in the hyposialylation of ␤1 and subsequently decreased cell migration. The restoration of GOLPH3 and the up-regulation of ST6GAL1 expression significantly rescued the sialylation of ␤1 integrin and cell migration. Furthermore, overexpression of ST6GAL1 efficiently rescued cell migration and AKT phosphorylation, which were blocked in the GOLPH3-knockdown cells (Fig. 6). Indeed, elevated levels of ST6GAL1 and ␣2,6-sialylation have been observed in several types of tumors, including colon cancer (18,52) and ovarian cancer (42,53). The expression levels of ST6GAL1 mRNA and enzyme activities are known to be particularly enhanced in metastatic tumors, which were promoted by the Ras oncogene (15,16). Consistently, in vitro cell culture studies have suggested that ST6GAL1 up-regulation contributes to cancer metastasis by regulating invasiveness and/or cell motility (18,54), as observed in this study. It is worth noting that the effects of GOLPH3-mediated sialylation on cell migration and cellular signaling could not be excluded from other target proteins, such as EGFR. These results prompted us to speculate that the regulation of sialylation could be a plausible mechanism for the oncogenic GOLPH3 in various cancer tissues.