Enhanced Epithelial-Mesenchymal Transition-like Phenotype in N-Acetylglucosaminyltransferase V Transgenic Mouse Skin Promotes Wound Healing*

N-Acetylglucosaminyltransferase V (GnT-V) catalyzes the β1,6 branching of N-acetylglucosamine on N-glycans. GnT-V expression is elevated during malignant transformation in various types of cancer. However, the mechanism by which GnT-V promotes cancer progression is unclear. To characterize the biological significance of GnT-V, we established GnT-V transgenic (Tg) mice, in which GnT-V is regulated by a β-actin promoter. No spontaneous cancer was detected in any organs of the GnT-V Tg mice. However, GnT-V expression was up-regulated in GnT-V Tg mouse skin, and cultured keratinocytes derived from these mice showed enhanced migration, which was associated with changes in E-cadherin localization and epithelial-mesenchymal transition (EMT). Further, EMT-associated factors snail, twist, and N-cadherin were up-regulated, and cutaneous wound healing was accelerated in vivo. We further investigated the detailed mechanisms of EMT by assessing EGF signaling and found up-regulated EGF receptor signaling in GnT-V Tg mouse keratinocytes. These findings indicate that GnT-V overexpression promotes EMT and keratinocyte migration in part through enhanced EGF receptor signaling.

Oligosaccharide structure changes are detected following birth, differentiation, and carcinogenesis (1), and these changes are regulated by glycosyltransferases. In particular, N-acetylglucosaminyltransferase V (GnT-V) 3 plays an important role in carcinogenesis and tumor metastasis (2). To characterize the detailed molecular mechanisms underlying GnT-V-related tumor metastasis, we and other groups succeeded in purifying and cloning GnT-V (3)(4)(5). In addition, we developed a sugar remodeling system of cancer cells and demonstrated the biological function of GnT-V in tumor metastasis through biochemical analysis of its target glycoproteins (6). Granovsky et al. (7) reported that mammary tumor growth and metastases induced by the polyomavirus middle T oncogene were considerably less in GnT-V-deficient mice than in littermate control mice. Cancer cells established from GnT-V-deficient mice showed lower cell growth and intracellular signaling than control mice because of aberrant glycosylation of growth factor receptors (8). Dennis et al. (9) also reported that sugar metabolism is critical to control the formation of ␤1,6 N-acetylglucosamine (GlcNAc), a product of GnT-V. In contrast, GnT-V is involved in negative regulation of T cell activation, leading to suppression of the autoimmune reaction (10). Recently, Mkhikian et al. (11) reported that genetic changes in glycosylation status for suppressing N-glycan branching are concerned with the incidence of multiple sclerosis. Our groups have studied biological functions of adhesion molecules such as cadherin and integrins in terms of N-glycan branching mediated by GnT-V and have found that GnT-V inhibits cell-cell/cell-matrix adhesion and promotes migration of cancer cells (12).
Although GnT-V is known to be up-regulated in the early phase of carcinogenesis in many cancers (13), it is unclear whether GnT-V regulates the late phase of cancer progression. In the case of colon cancer, GnT-V expression is associated with a poor prognosis (14). In contrast, low GnT-V levels are linked to a poor prognosis in lung and bladder carcinomas (15,16). These reports suggest that the role of GnT-V in cancer progression is organ-specific. Thus, we developed the GnT-V Tg mouse model to characterize the biological effects of GnT-V overexpression in normal tissues.
High GnT-V expression was observed in the skin of GnT-V Tg mice, and an epithelial-mesenchymal transition (EMT)-like phenotype was observed in cultured keratinocytes derived from these mice. EMT, which is characterized by the loss of epithelial adhesion and gain of mesenchymal features, is a fundamental biological process of embryonic development and cancer invasion/metastasis (17). In adults, EMT, which is driven by the cytokine bath generated by tissue injury, mediates the production of fibroblasts during inflammation and wound healing (18 -20). Re-epithelialization in wound healing involves the motility or migration of epithelial cells, and the migrating epithelial cells in wound margins acquire mesenchymal features and follow early stages of EMT (18).
In this study, we analyzed the EMT-like phenotypes of keratinocytes from GnT-V Tg mice and found that enhanced kerat-inocyte motility and cutaneous wound healing were mediated in part by aberrant EGF receptor signaling.

EXPERIMENTAL PROCEDURES
GnT-V Expression Vector-We used the expression vector pCAGGS, which contains a human ␤-actin promoter fused to a human ␤-globin gene fragment and polyadenylation signals (21). A construct containing the full-length GnT-V cDNA was constructed (5), as shown in Fig. 1.
Establishment of Transgenic Mice-Bromodomain factor 1 (BDF1) mice used in this study were purchased from SLC Co. (Shizuoka, Japan) and maintained under specific pathogen-free conditions. After microinjection of SalI-linearized Mgat5 expression vector into the pronuclei of BDF1-derived zygotes, the zygotes were introduced into the ampullae of pseudopregnant BDF1 mice, which gave birth to litters of 10 to 14 mice 21 days later. Newborn mice were genotyped by Southern blot analysis of BamHI-digested genomic DNA isolated from tail tissues, using a DNA fragment containing the Mgat5 gene-coding region as a probe. Transgene-positive F 1 mice were backcrossed to wild-type BDF1 mice to generate stable lines. Subsequently, transgene-positive mice were identified by PCR using genomic DNA from mice as templates and confirmed by Southern blot analysis with the probe described above. PCR conditions were 95°C for 1 min followed by 37 cycles of denaturation at 94°C for 30 s, annealing at 54°C for 30 s, and extension at 72°C for 50 s. The following human Mgat5 primers were used: 5Ј-GTGCTGGTTGTTGTGCTGTC-3Ј (sense) and 5Ј-CTT-GATTGCTTGGATCC-3Ј (antisense). Transgene-negative littermates were used as control mice. All mice were maintained in a specific pathogen-free facility at Osaka University. The Institutional Animal Care and Use Committee at Osaka University approved all procedures.
Cell Culture-Mouse keratinocytes were isolated and cultured as described previously (22). Full-thickness tail skin harvested from 12-week-old mice was treated with 4 mg/ml of dispase (Invitrogen) for 1 h at 37°C. The epidermis was then peeled from the dermis and trypsinized to prepare single-cell suspensions. Cells were incubated in human keratinocyte serum-free medium (DS Pharma Biomedical, Osaka, Japan) for 6 to 12 h at 37°C in 5% CO 2 to allow the cells to adhere to culture dishes precoated with type 1 collagen (Asahi Techno Glass, Funabashi, Japan). Non-adherent cells were washed with phosphate-buffered saline twice and then cultured for 2 to 3 days in human keratinocyte serum-free medium before use in experiments.
Assay of Enzyme Activities-Assay of GnT-V activity was carried out according to a previous report, with slight modifications (23). The assay employed 250 mM Mes buffer (pH 6.25) containing 400 mM UDP-GlcNAc, 20 mM EDTA, 400 mM GlcNAc 2 mg/ml BSA, and 1.0% Triton X-100. First, 0.5 l of 100 mM substrate was added to 6.5 l of this solution. Following this, 3.5 l of enzyme solution was added, and the mixture was incubated at 37°C for 4 h. The enzyme reactions were stopped by heating at 100°C for 5 min. The samples were added to 25 l water and centrifuged at 15,000 rpm for 15 min. Supernatants were applied to a TSKgel ODS-80TM column (4.6 ϫ 150 mm). Elution was performed at 55°C with a 0.1 M ammonium acetate buffer (pH 4.0) containing 1% n-butanol at a flow rate of 1.0 ml/min. The specific activity of GnT-V is expressed as pmol of GlcNAc transferred/h/mg of protein.
Immunohistochemistry and Immunofluorescence Staining of Skin Sections-Mouse dorsal skin samples were fixed in 10% formaldehyde for 24 h followed by paraffin embedding and microtome sectioning. Slides were then stained with H&E. For immunohistochemical analysis, sections were hydrated by passage through xylene and graded ethanols. After antigen retrieval for 10 min at 95°C in citrate buffer (pH 6), the slides were blocked with serum-free protein block (DakoCytomation) for 15 min and then incubated overnight at 4°C with the primary antibody mouse anti-E-cadherin (1:100 dilution, R&D Systems, Minneapolis, MN). After washing with TBS containing 0.05% Triton X-100, the slides were developed with the DAKO LSAPϩSystem-AP (DakoCytomation) and Dako ChemMate Envision kit/HRP(DAB) and then counterstained with hematoxylin. Rabbit and mouse IgG were used as the isotype controls.
Western Blot Analysis-Cell samples were solubilized at 4°C in TNE buffer consisting of 10 mM Tris-HCl (pH 7.8), 1% Nonidet P-40, 0.15 M NaCl, 1 mM EDTA, and protease inhibitor mixture (Wako, Osaka, Japan). For in vivo samples, skin specimens were crushed in liquid nitrogen and solubilized at 4°C in TNE buffer. The same amount of proteins (10 -30 g depend-ing on each experiment) was separated by SDS-PAGE and transferred onto polyvinylidene fluoride membranes (Bio-Rad). Nonspecific protein binding was blocked by incubation in 5% w/v nonfat milk powder in TBS-T (50 mM Tris-HCl pH 7.6), 150 mM NaCl, and 0.1% v/v Tween 20). The membranes were then incubated with rat anti-E-cadherin antibody (Sigma-Aldrich), rabbit anti-cytokeratin 5 antibody (Covance), mouse anti-␣-SMA antibody (DakoCytomation), rabbit anti-phospho Erk, rabbit anti-Erk (Cell Signaling Technology Inc., Beverly, MA), or mouse anti-GnT-V antibody 24D11 (provided by Fujirebio, Hachioji, Japan), each at 1:1000 dilution overnight at 4°C, or with mouse monoclonal anti-␤-actin (Sigma-Aldrich) at 1:5000 for 30 min at room temperature. The membranes were then washed three times in TBS-T (5 min each wash). Finally, the membranes were incubated with HRP-conjugated anti-rabbit, anti-mouse, or anti-rat antibodies (1:10,000 dilution) for 60 min at room temperature. Protein bands were detected by chemiluminescence using the ECL Plus kit (GE Healthcare). Band intensity was quantified with ImageJ software (National Institutes of Health, Bethesda, MD).
Lectin Blotting and Lectin Precipitation-For lectin blotting, membranes were blocked with 3% bovine serum albumin in TBS-T, followed by incubation with 10 g/ml biotinylated phytohemagglutinin-L 4 (L 4 -PHA) lectin (J-Oil Mills, Tokyo, Japan). Reactive bands were detected using the ECL Plus kit (GE Healthcare). For lectin precipitation, skin protein samples (2.7 mg) were incubated with 50 l L 4 -PHA lectin agarose (J-Oil Mills) for 2 h at 4°C and then centrifuged at 2000 rpm for 2 min. The lectin precipitates were washed two times with TNE buffer, and the bound proteins were boiled in 50 l 2ϫ SDS sample buffer, eluted, resolved by 6% SDS-PAGE, and finally transferred onto nitrocellulose membranes (GE Healthcare), as described previously (24). The membrane was incubated with rabbit anti-EGF-R antibody (1:500 dilution, Cell Signaling Technology) and probed with HRP-conjugated anti-rabbit, and proteins were detected by chemiluminescence.
In Vitro Migration Assay-Mouse keratinocytes were isolated as described previously, seeded into dishes precoated with type 1 collagen (1.3 ϫ 10 6 cells/3.5-cm dish), and grown to confluence. After serum starvation for 24 h, they were treated with 10 g/ml mitomycin C for 30 min to inhibit proliferation. A cell-free area was introduced by scraping the monolayer with a pipette tip. Cell migration to the cell-free area was evaluated in the presence or absence of EGF (10 ng/ml; R&D Systems) or the EGF-R inhibitor AG1478 (10 nM; LC Laboratories, Woburn, MA).
Statistical Analysis-The results are representative of at least three independent experiments. p values were calculated using a two-sided unpaired Student's t test.

Establishment of a Transgenic Mouse Line Expressing GnT-V-
The Mgat5 (GnT-V)-expressing construct used to generate mouse lines is shown in Fig. 1A. Expression of GnT-V protein  assessed by Western blot analysis was increased in many tissues of the transgenic mice, including skin, brain, pancreas, kidney, and liver (Fig. 1B). GnT-V activities evaluated by HPLC were almost consistent with the data of the Western blot analysis (Fig. 1C). To evaluate the oligosaccharide structures of skin glycoproteins, skin homogenates were assessed by lectin blot analysis with L 4 -PHA, which is known to react preferentially with ␤1,6 GlcNAc (25). Skin homogenates from GnT-V Tg mice showed higher reactivity to L 4 -PHA than those of wildtype mice (Fig. 1D). H&E staining of GnT-V Tg mouse skin did not reveal differences from the wild-type controls (Fig. 1E).

Enhanced Migration of GnT-V Tg Mouse Keratinocytes-Although GnT-V expression is up-regulated in various cancers,
GnT-V overexpression did not result in any spontaneous cancers in GnT-V Tg mice within the first year. However, the healing of minor wounds of the back skin made by fighting was markedly accelerated in GnT-V Tg mice compared with wildtype controls. We speculated that re-epithelialization might be enhanced in GnT-V Tg mice and therefore evaluated in vitro migration of GnT-V Tg mouse keratinocytes. The number of cells that migrated to a cell-free area (created by scraping the cell culture monolayer with a pipette tip) was significantly higher in GnT-V Tg keratinocytes than controls (Fig. 2, A-C). The cells of the migrating edge showed a spindle shape in GnT-V Tg mouse keratinocytes compared with wild-type control keratinocytes. The spindle-shaped cells expressed keratin 5, suggesting that they had the character of keratinocytes (Fig.  2D). A ratio of K i -67 positive cells was not different between wild-type and GnT-V Tg mouse keratinocytes, indicating that cell proliferation did not significantly differ between GnT-V Tg and wild-type keratinocytes (Fig. 2, E and F). In addition to enhanced cell migration, evaluation of primary keratinocytes in cultures derived from GnT-V Tg epidermis showed decreased expression of E-cadherin (Fig. 3A). The number of cells expressing N-cadherin was higher in GnT-V Tg keratinocytes compared with that of wild-type keratinocytes (Fig. 3B). Further, the number of ␣-SMA expressing cells was also increased in GnT-V Tg keratinocytes (Fig. 3C), and cells expressing both K5 and ␣-SMA epidermal and mesenchymal markers were also observed (D). Decreases in E-cadherin were confirmed by Western blot analysis (Fig. 3E). However, we were not able to detect N-cadherin by Western blot analysis (data not shown).
EMT-like Features Are Induced in GnT-V Tg Mouse Skin Tissue-Delocalization of E-cadherin, acquisition of mesenchymal characteristics, and increased cell mobility are features of the early stage of EMT (26). We evaluated these EMT features in GnT-V Tg mouse keratinocytes by first determining   EMT-related gene expression by real-time PCR analysis. Expression of the EMT-related transcription factors snail and twist were increased in GnT-V Tg mouse keratinocytes (Fig. 4). The switch from E-cadherin to N-cadherin results in loss of the epithelial phenotype and acquisition of the mesenchymal phenotype (27). We found that mRNA levels of N-cadherin was elevated in GnT-V Tg keratinocytes, although E-cadherin was not significantly reduced (Fig. 4). Taken together, these data demonstrate that EMT-like features were induced in GnT-V Tg mouse keratinocytes and may contribute to the enhanced migration of GnT-V Tg keratinocytes.
Aberrant Glycosylation of EGF-R Enhances Its Signaling in GnT-V Tg Mouse Keratinocytes-The autocrine/paracrine signaling of cytokines, such as EGF, TGF-␤, and TNF-␣, regulate EMT morphologic phases in keratinocytes (18,28,29). EGF family members are primary growth factors involved in re-epithelialization during cutaneous wound healing (30,31). High levels of ␤1,6 GlcNAc branching in EGF-Rs of GnT-V Tg mouse skin were inferred from the finding that EGF-R levels of GnT-V Tg mouse skin precipitated with L 4 -PHA lectin were significantly higher than those of controls (Fig. 5A). Glycosylation of EGF-R modulates EGF-R signaling (32). Therefore, we evaluated EGF-R signaling by determining the level of ERK phosphorylation induced by EGF (10 ng/ml). As expected, prolonged phosphorylation of Erk was observed after EGF treatment of GnT-V Tg keratinocytes (Fig. 5B), and migration was enhanced by EGF in these cells (C). These findings suggest that increased ␤1,6 GlcNAc branching on EGF-R induced by GnT-V overexpression modulates EGF-R signaling, thereby altering EMT features in the skin.

EGF-R Inhibitor Attenuates Migration of GnT-V Tg
Keratinocytes-To confirm the significance of EGF-R signaling in the EMT-like features of GnT-V Tg mice, we evaluated the effects of the EGF-R inhibitor AG1478 on keratinocyte migration in vitro. We found that AG1478 attenuated migration in GnT-V Tg keratinocytes (Fig. 5D). These results showed that EGF-R signaling was involved in the enhanced migration of GnT-V Tg keratinocytes.
Aberrant Glycosylation Was Increased on E-cadherin in GnT-V Tg Mouse Keratinocytes-We next examined whether or not glycosylation of E-cadherin was altered in GnT-V Tg mouse keratinocytes. Cell lysate of wild-type and GnT-V Tg mouse keratinocytes was immunoprecipitated with anti-E-cadherin antibodies, and the binding of immunoprecipitants to L 4 -PHA was evaluated by lectin blot analysis. Although levels of total immunoprecipitated E-cadherin were lower in GnT-V Tg mouse keratinocytes, almost the same intensity of bands binding to L 4 PHA was observed between GnT-V Tg and wild-type mouse keratinocytes, suggesting that aberrant glycosylation on E-cadherin was increased in GnT-V Tg mouse keratinocytes (Fig. 5E).
Wound Healing Was Accelerated in GnT-V Tg Mice-To evaluate the EMT-like phenotype and the enhanced migration of GnT-V Tg mouse keratinocytes in vivo, we performed a cutaneous wound healing assay using 8-mm round wounds on the backs of GnT-V Tg and control mice. The wound areas in GnT-V Tg mice were significantly lower than controls on days 2, 4, 6, 8, and 10 ( Fig. 6, A and B). Mean wound closure was 16.0 Ϯ 1.81 days in control mice and 14.4 Ϯ 1.91 days in GnT-V Tg mice (Fig. 6C). Notably, re-epithelialization was more rapid in GnT-V Tg mice (Fig. 6D). These results demonstrate that re-epithelialization of cutaneous wounds was enhanced in GnT-V Tg mice because of the enhanced EMT-like phenotype.

DISCUSSION
Recently, EMT has become a focus of cancer research, as molecular and morphologic features of EMT are found to correlate with poor histological differentiation, loss of tissue integrity, and metastasis (33). Characteristics of EMT in vitro include increased expression of mesenchymal factors (collagen I and vimentin), cadherin switch, loss of epithelial markers, spindle-shape morphology, increased migratory capacity, and resistance to apoptotic stimuli (17).
In this study, we detected EMT-like features in the skin of GnT-V Tg mice, including expression of mesenchymal factors, spindle-shape morphology, and increased migration (Fig. 2). Epithelial wound closure, dermal repair, and angiogenesis are steps in the repair of cutaneous wounds (31,34). Motility and FIGURE 6. Enhanced wound healing was observed in GnT-V Tg mice. An 8-mm punch biopsy of the skin was obtained from the backs of 16-week-old male wild-type and GnT-V Tg mice, and wound closure was monitored (wild-type mice, n ϭ 13, Tg mice, n ϭ 17). A, reduction of the wound area on days 2, 4, 6, 8, and 10. *, p Ͻ 0.05. B, macroscopic view of wound healing on day 8 after biopsy. C, time required for wound closure (days, mean Ϯ S.D.). D, expanded photograph of re-epithelialization in wild-type and GnT-V Tg mice on day 6. Dotted lines show the re-epithelialized edge of the epidermis. Bars indicate mean re-epithelialization (distance) on day 6. *, p Ͻ 0.05. migration of epithelial cells, which are key steps in re-epithelialization, follow the early stage phenomena of EMT (18,35). Cutaneous wound healing in vivo and keratinocyte mobility in vitro was accelerated in GnT-V Tg mice because GnT-V overexpression promoted EMT in the epidermis of transgenic mice (Fig. 4). This result was consistent with data of a previous in vitro study that overexpression of GnT-V in Mv1Lu cells enhanced the migration to the scratch wound (36).
GnT-V expression is associated with a poor prognosis in breast cancer (37), colon cancer (14), and endometrial cancer (38). Consistent with these studies, we found that GnT-V was up-regulated in some cases of poorly differentiated invasive squamous skin carcinoma 4 . These results suggest that GnT-Vassociated EMT is involved in the progression of these carcinomas.
EMT features in cutaneous wound healing differ from the features of cancer metastasis because epithelial cells involved in wound healing that acquire mobility and mesenchymal phenotypes return to the epithelial phenotype. In GnT-V Tg mice, the early phase phenomena of EMT and enhanced re-epithelialization were observed, but spontaneous carcinoma did not occur during the 1-year study period, suggesting that GnT-V overexpression in the skin is not sufficient to induce cancer development. Gene mutation(s) and oncogene activation appear to be required for spontaneous carcinogenesis. Thus, GnT-V overexpression in premalignant conditions may produce results that differ from those of this study. Experimental carcinogenesis in our GnT-V Tg model may be useful to answer this problem in the future.
EGF-R signaling essential for keratinocyte migration and proliferation in EMT is predominantly regulated by autocrine EGF-R activation (39,40). EGF-R-mediated proliferation and migration of keratinocytes also appears to be crucial in wound healing. Keratinocytes play a central role in wound repair as a source of growth factors (26,39). Previous studies reported that GnT-V-deficient mammary tumor cells showed reduced EGF-R signaling through decreased galectin-3 binding to polylactosamine structures and rapid internalization of the receptors from the cell surface (41). Mammary tumor cells from GnT-V knockout mice are insensitive to EGF, fibroblast growth factor, and TGF-␤ (8). Down-regulation of ␤1,6 GlcNAc branching by GnT-III overexpression also decreased EGF-R signaling (26). As EGF-R signaling was increased in GnT-V Tg mouse keratinocytes, we hypothesized that overexpression of GnT-V promotes an EMT-like phenotype in wound healing in part by modulated EGF-R signaling.
We also assume that modification of E-cadherin N-glycans by GnT-V also played a role in keratinocyte migration, as posttranscriptional modification of E-cadherin by GnT-V is reported to delocalize E-cadherin to cytoplasm (42,43). Because expression of E-cadherin was not altered at mRNA level but was decreased at protein level in GnT-V Tg mouse keratinocytes, we consider that an increase in aberrant glycosylation of E-cadherin might down-regulate the expression of E-cadherin at the protein level. In our previous reports, the reverse phenomenon was observed in mouse melanoma cells transfected with the N-acetylglucosaminyltransferase III (GnT-III) gene (44). Because GnT-III suppresses a reaction of GnT-V and prolongs the half-life of E-cadherin, the instability of E-cadherin modulated by GnT-V in this study seems to be compatible with our previous data. Modulation of her-2-mediated signaling pathways by GnT-V might also affect GnT-V Tg mouse keratinocytes, as it is known to regulate the proportion of tumor-initiating cells (45).
In conclusion, we found that EMT features in GnT-V Tg mice that contributed to keratinocyte mobility, and cutaneous wound healing was mediated in part by up-regulated EGF receptor signaling. These findings suggest that GnT-V overexpression in keratinocytes induces the early phase of malignant transformation. Future experiments involving carcinogen induction or mating GnT-V Tg mice with other carcinogenic mice may be useful to test this hypothesis.