Fibroblast growth factor 2 accelerates the epithelial–mesenchymal transition in keratinocytes during wound healing process

In the wound healing process, the morphology of keratinocytes at the wound edge temporarily changes to a spindle morphology, which is thought to occur due to an epithelial–mesenchymal transition (EMT). Fibroblast growth factor (FGF) 2, also called basic FGF, has the potential to accelerate wound closure by activating vascular endothelial cells and fibroblasts. We examined the effects of FGF2 on keratinocyte morphology and EMT in wounded skin. Histological examination of murine wounds treated with FGF2 revealed that wound edge keratinocytes formed thickened and multilayered epithelia. In addition, we detected wound edge keratinocytes migrating individually toward the wound center. These migrating keratinocytes exhibited not only spindle morphology but also down-regulated E-cadherin and up-regulated vimentin expression, which is characteristic of EMT. In FGF2-treated wounds, a PCR array revealed the upregulation of genes related to EMT, including transforming growth factor (TGF) signaling. Further, FGF2-treated wound edge keratinocytes expressed EMT-associated transcription factors, including Snai2, and showed translocation of β-catenin from the cell membrane to the cytoplasm/nucleus. However, in vitro examination of keratinocytes revealed that FGF2 alone did not activate EMT in keratinocytes, but that FGF2 might promote EMT in combination with TGFβ1. These findings suggest that FGF2 treatment of wounds could promote keratinocyte EMT, accelerating wound closure.

Snai1 superfamily of zinc finger transcriptional repressors, is the most investigated molecule in wound healing EMT, and is thought to be the main EMT-inducing transcription factor in cutaneous wound healing 6,12,13,17,18 .
Fibroblast growth factor (FGF) 2, also called basic FGF, is a member of a large FGF family of structurally related proteins that bind heparin sulfate and modulate the growth, differentiation, migration and survival of a wide variety of cell types 19 . FGF2 strongly activates not only fibroblasts but also other mesoderm-derived cells, including vascular endothelial and smooth muscle cells, osteoblasts and chondrocytes 20 . Administration of recombinant FGF2 to skin wounds accelerates acute and chronic wound healing [21][22][23] . Therefore, topical recombinant FGF2 has been approved in Japan for the treatment of skin ulcer since 2001. Local FGF2 administration also has an anti-fibrotic effect for the wound to antagonize myofibroblast differentiation and dampen fibrosis which has been evidenced by decreased SMA 24 and fibronection in the wound tissue 25 . In addition, FGF2 is considered to accelerate reepithelization 26 , which is especially prevalent in an epidermis-defect wound model 27 . However, most in vitro studies have reported that FGF2 is less active in keratinocytes that are derived from primitive ectoderm, relative to mesoderm-derived cells 28 .
In the present study, we detected wound edge migrating and spindle keratinocytes in a mouse cutaneous wound model, which were more prevalent in FGF2-treated wounds. We tested the hypothesis that local wound FGF2 treatment would cause wound edge keratinocytes to undergo EMT, facilitating rapid migration on the wound bed to speed wound closure. Further, we characterized FGF2-induced temporal changes in keratinocytes, focusing on the molecular mechanisms of EMT.

FGF2 accelerated wound healing, accompanied by individually migrating keratinocytes.
To investigate the effects of FGF2 on cutaneous wound healing, 6 mm biopsy punch wounds were generated on the dorsal regions of C57BL/6 mice and wound healing was observed. On days 2 and 4, FGF2-treated wounds closed more rapidly than vehicle-treated wounds (Fig. 1a). The average period required for wound closure was significantly shorter in the FGF2-treated group (7.8 ± 0.2 days) than in the control group (9.8 ± 0.3 days, p < 0.01; Fig. 1b). In addition, histological examination 4 days after wounding revealed thickening on the edge of the epidermal layer in FGF2-treated wounds (Fig. 1c). The numbers of keratinocytes layer of wound edge were increased in FGF2-treated wounds, with an average of 12.5 ± 2.5 keratinocytes in the FGF2-treated group and 7.0 ± 1.5 keratinocytes in the control group (p < 0.01; Fig. 1d). At day 4 after induction of skin wounds, histological examination revealed that keratinocytes in FGF2-treated wounds protruded from the surface, and that spindle-shaped keratinocytes migrated from the wound edge into fibrin clots or granulation tissue (Fig. 1e). Immunofluorescence staining with an anti-pan-keratin antibody confirmed the migrating cells to be keratinocytes (Fig. 1f). The number of migrating keratinocytes was significantly higher in the FGF2-treated group (4.3 ± 1.2 cells) than in the control group (0.3 ± 0.2 cells, p < 0.01; Fig. 1g). Thus, murine FGF2-treated wounds exhibited rapid wound closure, accompanied by epithelial proliferation at the wound edge. Additionally, the characteristic spindle morphology and individual migration observed in FGF2-treated wounds suggested that wound edge keratinocytes had acquired a mesenchymal phenotype.

Immunofluorescent evaluation of EMT-related cell morphological components.
To determine if the observed individually migrating cells had acquired mesenchymal cell characteristics, we immunostained FGF2-treated wounds 4 days after wounding. Cell membrane E-cadherin expression was reduced in wound edge keratinocytes both in FGF2 treated wounds and in vehicle-treated wounds. The down-regulated E-cadherin was also seen in keratinocytes detached from the wound edge in the fibrin clot (Fig. 2a). Interestingly, the wound edge and individually migrating keratinocytes in the FGF2-treated wound co-expressed the intermediate filaments cytokeratin and vimentin (Fig. 2b). Among sections of FGF2-treated wounds on day 4, we detected spindleshaped monolayer keratinocytes on the portion of granulation tissue referred to as the epithelial tongue (Fig. 2c). These migrating keratinocytes were strongly positive for vimentin and cytokeratin. In addition, cell membrane E-cadherin was nearly undetectable (Fig. 2d). The number of vimentin and cytokeratin co-expression cells in the wound edge and migrating keratinocytes were significantly higher in FGF2 treated wounds (30.2 ± 5.3 cells) than in the control group (5.2 ± 2.6 cells, p < 0.01; Fig. 2e). Thus, these findings suggested that wound edge and migrating keratinocytes decreased E-cadherin regardless of the FGF2 application or not. In addition, those keratinocytes observed in FGF2-treated wounds acquired typical morphological and molecular features of EMT, which were characterized by not only decreased E-cadherin but also increased vimentin.
PCR array analysis of wound tissue EMT-associated molecules. A tissue PCR array of 84 EMTassociated components (See Supplementary Data. S2 online) was performed using RNA extracted from fullthickness mouse wounds on day 4, treated with FGF2 (n = 3) or PBS (n = 3). Out of targets, 34 transcripts were at least 1.5-fold upregulated, and 11 were at least 1.5-fold down-regulated in FGF2-treated wounds compared to control wounds. Genes were categorized into five groups, including genes typically upregulated during EMT, cell growth and proliferation genes, cell migration and motility genes, genes related to TGF/bone morphogenetic protein (BMP) signaling, and genes related to Wnt signaling. Of these gene categories, the 34 upregulated genes in the in FGF-treated group included ten genes in the EMT category, nine genes in the cell growth and proliferation category, five genes in the cell migration and motility category, four genes in the TGF/BMP signaling category and three genes in the Wnt signaling category (Fig. 3). Fewer genes in each category were downregulated in the FGF2 group. To confirm the expression of major EMT-associated genes, we selected representative genes that are commonly investigated in the study of EMT from the PCR array data. Individual results from PCR array revealed changes in mRNA levels of TGFβ1 (2.79 ± 0.34 fold: p < 0.01), TGFβ2 (1. 18 (Fig. 4). Taken together, these findings suggested that EMT associated components, including transcriptional regulators and cytokines in the TGF/ BMP and Wnt signaling axes, were activated in FGF2-treated wound tissue. TGFβ1, a potent inducer of EMT, was significantly upregulated in FGF2-treated wound tissue, suggesting that TGFβ1 could be a key regulator of FGF2-mediated EMT enhancement.
Immunofluorescent staining of EMT-associated transcription factors. We performed tissue immunofluorescent staining of the non-affected epidermis (normal skin), vehicle-treated wound (control), and FGF2-treated wound at day 4 after wounding to evaluate expression and localization of EMT-associated factors at wound edge keratinocytes. As a representative result (Fig. 5), Snai1 was not present in normal skin, control and FGF2 treated-wound. Snai2 was robustly expressed in the nucleus of wound edge keratinocytes in the FGF2 www.nature.com/scientificreports/ group, faintly expressed in a few keratinocytes in control and not expressed in normal skin. Twist was detected in the cytoplasm of wound edge keratinocytes in the FGF2 group, but almost not detected in the normal skin and in the control group. β-catenin comprises the cell attachment structure of keratinocytes. In FGF2-treated wounds, some wound edge keratinocytes showed activation of β-catenin signaling evidenced by translocation of β-catenin from the cell membrane to the cytoplasm or nucleus, which was slightly seen in the control mock treated group. The immunofluorescent staining pattern of Notch1 did not differ among normal epidermis, control wounds and FGF2-treated wounds. Taken together, these findings suggest that keratinocyte EMT in FGF2treated wounds might be regulated by transcription factors such as Snai2, Twist and β-catenin.
In vitro assay of cultured keratinocytes treated with FGF2 and TGFβ1. In Figure 3. Tissue PCR array analysis about EMT associated components in murine wounds. Eighty-four EMT associated components were examined in RNA extracted from whole wounds at day 4 treated with vehicle or FGF2 (n = 3/group). Genes were categorized into five groups, including genes typically upregulated during EMT, cell growth and proliferation genes, cell migration and motility genes, genes related to TGF/bone morphogenetic protein (BMP) signaling, and genes related to Wnt signaling. Genes indicated with colored bars were up or down-regulated > 1.5 fold change. www.nature.com/scientificreports/ transcription factors in wound edge keratinocytes. As TGFβ1 is a major inducer of EMT 12 , we next used an in vitro assay, in which normal human epidermal keratinocytes (NHEKs) and a keratinocyte cell line were treated with TGFβ1 and FGF2 solo-stimulation and TGFβ1 and FGF2 co-stimulation (TGFβ1 + FGF2) to determine whether FGF2 contributed to the EMT on keratinocyte. In the experiment using NHEKs, TGFβ1 and TGFβ1 + FGF2 stimulation showed disseminated proliferation of cells that implied EMT in contrast to the cobblestone pattern of control and FGF2 treated cells 48 h after stimulation (Fig. 6a). mRNA expression of E-cadherin, vimentin, and Snai2 was highly and significantly upregulated in TGFβ1 and TGFβ1/FGF2. Whereas treatment of cells with FGF2 alone did not alter the expression of E-cadherin, vimentin, and Snai2. TGFβ1 alone significantly upregulates the mRNA of these genes, however, co-treatment of keratinocytes with TGFβ1 and FGF2 further enhances the expression of only E-cadherin (Fig. 6b). A keratinocyte cell line, HaCaT cells, gradually showed not only disseminated pattern but also spindle morphology with TGFβ1 and TGFβ1 + FGF2 treatment in the time course of cell culture (Fig. 7a). In contrast, morphology was unaffected with FGF2 alone. mRNA expression of E-cadherin, vimentin and Snai2 was highly and significantly upregulated in TGFβ1. Vimentin and Snai2 were further upregulated by TGFβ1 + FGF2 co-treatment than TGFβ1 solo treatment especially at the time point of 72 h (Fig. 7b). As these results using HaCaT cells revealed EMT with TGFβ1 and might imply enhancement of the TGFβ1-induced EMT with FGF2, we examined the mRNA expression of other EMT associated molecules (Fig. 7c). Snai1 increased with TGFβ1 but suppressed with FGF2 and TGFβ1 + FGF2 at 72 h. Twist1 was increased with TGFβ1 at 48 and 72 h, and TGFβ1 + FGF2 at 48 h. β-catenin was similarly upregulated by TGFβ1 and TGFβ1 + FGF2 treatment at both time points. Among EMT-inducing transcription factors, Notch1 was significantly suppressed by TGFβ1 treatment and TGFβ1 + FGF2 co-treatment. Treatment with FGF2 alone had little effect on the expression of EMT-related transcription factors in NHEK and HaCaT cells. However, when we see the expression levels of vimentin and Snai2 in TGFβ1-stimulated HaCaT cells with or without FGF2, FGF2 might affect the synergistic effect of these genes under TGFβ1 stimulation. To resolve the mechanism, we measured mRNA of FGFR1 which was one of the FGF2 receptors. As a result, FGFR1 showed a significant increase with TGFβ1 in comparison with control and FGF2 at 48 h in HaCaT cells (Fig. 7d). Taken together, NHEKs and HaCaT cells treated with TGFβ1 showed EMT but not with FGF2 alone. In HaCaT cells, some EMT markers like vimentin and Snai2 were further upregulated by TGFβ1 + FGF2 co-treatment relative to TGFβ1 treatment alone, suggesting that FGF2 might enhance TGFβ1-mediated EMT in keratinocytes via agonizing TGFβ1-upregulated receptors like FGFR1 by expression of FGF2 recognizable receptors such as FGFR1 by TGFβ1.

Discussion
We investigated the FGF2 function in the wound healing process, focusing specifically on keratinocytes. As reported in several papers before 21,23,26,27,29 , FGF2 treated wound closed faster than vehicle-treated wounds. In the epidermis of wound edge, in addition to epidermal hypertrophy, spindle-shaped and migrating keratinocytes were observed. These characteristic cells revealed to express intermediate filaments, not only cytokeratin but also vimentin which usually express in mesenchymal cells and to reduce their expression of E-cadherin, suggesting an EMT. In order to evaluate the tissue molecular condition in the FGF2 treated wound, we performed a PCR array using the wound tissue which showed that the tissue environment tended to initiate EMT as well as  www.nature.com/scientificreports/ www.nature.com/scientificreports/ cell growth, proliferation, cell migration and motility, and accelerate TGF/BMP signaling and WNT signaling. EMT associated transcriptional factors such as Snai2, Twist and β-catenin were also expressed in the wound edge keratinocytes. In in vitro examination, FGF2 treatment showed no influence on NHEKs and HaCaT cells, but FGF2 might have additional EMT effects to cells when co-stimulating with TGFβ. In partial epidermal EMT during wound healing, the cell-cell adhesion of keratinocytes is characterized by diminished desmosome components and adherens junction components, such as E-cadherin 5,6,30,31 . In contrast to carcinomas, upregulation of vimentin in wound healing-related EMT is still unclear 13 . Only one report demonstrated vimentin expression in migrating epithelial tongues of acute wounds and hypertrophic scars 32 . The present study identified vimentin expression in FGF2-treated wounds, especially in keratinocytes observed on the epithelial tongue, indicating that vimentin expression in keratinocytes during wound healing EMT could require specific conditions, such as tissue activation with FGF2. A study has revealed that vimentin-deficient wounds exhibited loss of EMT-like keratinocyte activation, limited keratinization and slow reepithelization 33 , demonstrating the importance of vimentin-positive keratinocytes in the reepithelization process.
Among EMT-associated transcription factors, several reports support the functional involvement of Snai2 in cutaneous wound healing 6,17,18 . In the present study, Snai2 expression was clearly observed in wound edge keratinocytes treated with FGF2 in vivo. Additionally, TGFβ1 induced Snai2 mRNA expression by which was significantly upregulated by co-treatment with FGF2 in HaCaT cells. A prior study identified that Snai2-null mice presented non-healing cutaneous ulcers in response to ultraviolet radiation 6,34,35 . Upregulation of Snai2 in the nuclei of keratinocytes at the edges of FGF2-treated wounds 36,37 supported the importance of Snai2 in keratinocyte migration and tissue regeneration. Other EMT-associated transcription factors such as Twist 38 and β-catenin 36,37 have been reported to express in keratinocytes at wound-healing process, as well as in tumor cells as hallmarks of EMT 14 .
The Snai1 family, including Snai1 and Snai2, regulates EMT associated gene transcription downstream of TGF signaling 18,39 . FGF2-treated wounds exhibited 3.05-fold upregulation of TGFβ1 in comparison to vehicletreated wounds. This result indicated that TGFβ1 could have provoked EMT as a messenger of FGF2 in the wound tissue. Translocation of β-catenin from the cell membrane to the cytoplasm or nucleus was observed in the wound edge keratinocytes of wounds treated with FGF2. E-cadherin not only connects keratinocyte membranes, but also sequesters β-catenin to the cell membrane to suppress Wnt-β-catenin signaling, which regulates key migration events of EMT 40 . In keratinocytes, this phenomenon is thought to participate in the wound healing process 41 . Notch signaling has various functions, including epidermal differentiation 41 and EMT 11 . Our PCR array results revealed greater expression of Notch1 mRNA in FGF2-treated wound tissue than in control www.nature.com/scientificreports/ wound tissue, however Notch1 was only marginally induced in wound edge keratinocytes in either group and also down-regulated in in vitro HaCaT model treated with TGFβ1 and TGFβ1 + FGF2. Notch1 is essential for epidermal proliferation and differentiation 42 . A recent study revealed that Notch1 expression is down-regulated in epidermal keratinocytes at the regeneration phase in the wound healing process 43 . From these facts, Notch1 would contribute to suppress epidermal differentiation but not influence the EMT in wound edge keratinocytes even when stimulating with topical FGF2. The significant up-regulation of Notch1 at the PCR array data in FGF2 treated wound tissue might be due to overall activation of cells especially in the regenerated dermis. Taken together, our results suggest that when FGF2 is applied to wounds during the acute phase, increased TGFβ1 may stimulate EMT signaling in wound edge keratinocytes, upregulating EMT-associated transcription factors such as Snai2. Finally, EMT, which is characterized by diminished E-cadherin and expression of vimentin, could occur, supporting wound healing. Although FGF2 accelerates reepithelization in the wound healing process 26,27 , the direct effects of FGF2 on epidermal keratinocytes remain controversial. A small number of studies have suggested that direct FGF2 stimulation activates keratinocytes. In vitro studies using NHEKs demonstrated that FGF2 could stimulate keratinocyte proliferation 44 . Nakamizo et al. demonstrated FGF2 promoted NHEKs proliferation, which was inversely dependent on cell density, as cells in high-density culture exhibited decreased FGF receptor expression 45 . In the present study, we demonstrated that FGF2 treatment alone had little impact on EMT and cell morphology in NHEKs and HaCaT cells. Intriguingly, our results on HaCaT cells also suggested that TGFβ1/FGF2 co-treatment upregulated EMT-associated genes more robustly than TGFβ1 treatment alone. Shirakihara et al. suggested that TGFβ and FGF2 cooperatively initiate EMT using NMuMG cells, a mouse mammary epithelial cell line. When NMuMG cells were co-stimulated with TGFβ and FGF2, TGFβ regulated isoform switching of FGF receptors from FGFR2IIIb into a FGF2-recognizable FGF receptor, FGFR1IIIc. Consequently, FGF2 co-treatment increased the migratory capacity of NMuMG cells, and induced morphological changes consistent with EMT relative to TGFβ treatment alone 46 . Our HaCaT model also might suggest a similar mechanism, by which TGFβ might initiate FGF receptor isoform switching into FGFR1 to allow FGF2 recognition, further promoting EMT.
In the present study, several limitations exist. In the in vivo wound healing model mainly documented in Fig. 1, additional factors modulating the acceleration of wound healing should be considered. Skin contraction resulting in faster healing, with temporarily increased myofibroblast is the main player 47 . TGFβ is known to increase differentiation of fibroblasts into myofibroblasts 48,49 , and thus, local administration of FGF2 which induce TGFβ mRNA in the tissue may indirectly reduce time to healing. On the other hand, local administration of FGF2 is also reported to suppress differentiation of fibroblasts to myofibroblasts in vivo 50,51 . Taken together, topical administration of FGF2 was revealed to shorten the period of wound healing despite suppression of myofibroblast differentiation. The mechanism responsible for FGF2 accelerating wound healing without skin contraction can be speculated to include neovascularization, modulated collagen and matrix deposition, as well as keratinocyte EMT. Various cell types may participate in these processes, but the current study focused on wound edge keratinocytes, representing a difficult-to-quantify site, both in morphology and molecular expression.
In summary, the results obtained in the present study suggest that topical application of FGF2 on cutaneous wound upregulates tissue mRNA of TGFβ1 and other EMT associated molecules, accelerates EMT of wound edge keratinocytes, and might contribute to reepithelization during new tissue formation.

Materials and methods
Mouse wound healing experiment. All experiments were approved by the committee for Animal Experimentation of Nagasaki University Graduate School of Biomedical Sciences (approval number 1710261420). All methods were carried out in accordance with relevant guidelines and regulations. Male C57BL6/J mice, 8-12 weeks old, were used for experiments and analysis. The dorsal skin was shaved and and two full thickness excisional wounds were created with a 6 mm sterile biopsy punch on the left and right dorsal skin. Recombinant FGF2 (trafermin; Kaken, Japan) diluted to 100 μg/ml (approved concentration for skin ulcer on human 29 ) with phosphate buffered saline (PBS) was prepared. The FGF2 solution (50 μl) was dropped on the right wounds, and PBS (50 μl) was dropped on the left wounds as a control, and then covered with film dressing. For the following days, daily treatment with 50 μl PBS or FGF2 solution and dressing changes were performed. The day of macroscopic wound closure was defined as the first day with no exudate fluid and no scab (n = 6, respectively).
Histology. Skin samples day 4 after wounding were obtained from seven murine and fixed in formalin and paraffin-embedded. Sections were stained with hematoxylin and eosin staining. Morphologic characteristics of wound samples were assessed by light microscopy. We chose one section whose both sides of epidermal edges had been clearly found from individual samples and counted numbers of cell layers at the most thickened site and migrating keratinocytes of the bilateral epidermis wound edge (n = 14, individually).  Real-time reverse transcription PCR. Total HaCaT cell RNA was isolated using a RNeasy kit (Qiagen).
Total RNA from each sample was reverse-transcribed into cDNA and analyzed with primers (See Supplementary Data. S1 online) and an ABI Prism 7000 sequence detector (Applied Biosystems, US) using a real-time PCR quantification method according to the manufacturer's instructions and description in the past report 55 . Sequence-specific primers and probes were designed by pre-developed TaqMan assay reagents or Assay-On-Demand (Applied Biosystems). Relative expression of real-time PCR products was determined using the ΔΔC T method to compare target gene and GAPDH mRNA expression.

Statistical analyses.
The results are given as means ± standard error of the mean. Statistical analyses were performed using GraphPad Prism 8 (GraphPad, San Diego, CA, USA). When two groups were analyzed, the data were performed using Mann-Whitney U test. When three or more groups were analyzed, the data were analyzed by ANOVA with a post hoc Tukey examination. p < 0.05 was considered statistically significant.
Received: 10 December 2019; Accepted: 14 October 2020 Figure 7. In vitro experiment using HaCaT cells to compare among normal culture condition, TGFβ1 solo-stimulation, FGF2 solo-stimulation, and TGFβ1 and FGF2 co-stimulation (TGFβ1 + FGF2). (a) HaCaT cells gradually showed not only the disseminated pattern but also spindle morphology with TGFβ1 and TGFβ1 + FGF2 treatment in the time course of cell culture from 24 to 72 h. Higher magnification images are at the corners. Scale bar: 20 μm (b) Realtime RT-PCR analysis of HaCaT cells. The Y-axis represents the relative ratio compared with CTL. mRNA expression of E-cadherin, vimentin and Snai2 was highly and significantly upregulated in TGFβ1. Vimentin and Snai2 were further upregulated by TGFβ1 + FGF2 co-treatment than