Hepatocyte growth factor/scatter factor inhibits UVB-induced apoptosis of human keratinocytes but not of keratinocyte-derived cell lines via the phosphatidylinositol 3-kinase/AKT pathway.

Acute irreparable UV-induced DNA damage leads to apoptosis of epidermal keratinocytes (KC) and the formation of sunburn cells, whereas less severely damaged cells survive but harbor the potential of tumor formation. Here we report that hepatocyte growth factor/scatter factor (HGF/SF) prevents UVB-induced apoptosis in primary KC cultured in vitro. When we analyzed the signaling pathways initiated by the HGF/SF receptor c-met, we found that the phosphatidylinositol (PI) 3-kinase and its downstream-element AKT and the mitogen-activated protein (MAP) kinase were activated. Inhibition of PI 3-kinase led to a complete abrogation of the anti-apoptotic effect of HGF/SF, whereas blockade of the MAP kinase pathway had no effect. In contrast to the observation with primary KC, HGF/SF could not enhance survival after UVB irradiation of HaCaT and A431 cell lines, despite the fact that in these cells the PI 3-kinase and MAP kinase pathways were also activated by HGF/SF. Cell cycle analysis of KC revealed a G(2)/M arrest after UVB irradiation and a complete loss of proliferating cells. Because HGF/SF in the skin is produced by dermal fibroblasts, our findings suggest that the HGF/SF-mediated rescue of KC from apoptosis represents an important paracrine loop by which UVB-damaged KC can be kept alive to maintain the epidermal barrier function but cannot further proliferate, thereby preventing the induction of epithelial skin tumors.

HGF/SF 1 derived from mesenchymal cells is a multifunctional cytokine that has been shown to have a variety of effects on cells of different origin. It was first identified as a potent mitogen for hepatocytes (1,2), and more recently it was shown to promote cell motility and proliferation of KC, melanocytes, and kidney epithelial cells (for review see Ref. 3). Furthermore, HGF/SF is able to induce scattering of cells (4) and their invasion into extracellular matrix (5), thereby promoting tumor metastasis (6,7). Recently HGF/SF has been shown to promote or inhibit apoptosis depending on the cell type. For example, in the renal epithelial cell line HKC, HGF/SF acts as a survival factor after serum withdrawal (8,9); it inhibits hepatocyte apoptosis in Fas-induced fulminant hepatic failure in a mouse model (10) but is able to induce apoptosis of sarcoma 180 cells (11).
The receptor for HGF/SF is a heterodimeric tyrosine kinase encoded by the c-met proto-oncogene, consisting of a 50-kDa extracellular ␣-subunit and a 145-kDa transmembrane ␤-subunit (12). c-met signaling is mediated by autophosphorylation on Tyr 1334 and Tyr 1335 of the ␤-subunit, which leads to a strong up-regulation of its kinase activity (13). As a result, a multifunctional docking site for adaptor molecules on the C-terminal tail of the ␤-subunit becomes phosphorylated at the tyrosine residues on position Tyr 1349 and Tyr 1356 . Their phosphorylation leads to interaction of c-met with several cytoplasmic signal transducers. This occurs either directly, such as with the PI 3-kinase, or indirectly via molecular adapters such as Grb2 (14), Shc (15), or Gab1 (7), activating the MAP kinase (16) and the STAT-3 pathways (17). The MAP kinase pathway has been shown to be responsible for cell growth (14), whereas the phosphorylation of STAT-3 and the resulting nuclear signaling is required for triggering differentiation for branching morphogenesis (17). The PI 3-kinase pathway is responsible for cell scattering by inducing the loss of intercellular junctions and cell migration (18). This pathway has recently been shown to also be involved in the anti-apoptotic activity of HGF/SF in renal tubular epithelial cell line (19), in hepatocytes (20), and in NIH 3T3 fibroblasts (21).
For the skin, UV irradiation is the most important DNA damaging stimulus and represents the major risk factor for the development of epithelial skin tumors (22). Whereas mild UVinduced damage induces DNA repair, severe UV exposure leads to irrepairable DNA damage resulting in KC apoptosis and the formation of sunburn cells (23,24). It has been suggested that this UV-induced apoptosis contributes to the homeostasis of the epidermis and helps to prevent skin cancer by preferentially eliminating DNA-damaged KC (25). However, because a substantial loss of keratinocytes would result in a life-threatening damage of the skin barrier function, we are interested in mechanisms that counteract UV-induced KC apoptosis. In the present study we demonstrate that HGF/SF inhibits UV-induced apoptosis of KC and arrests them irreversibly in the G 2 /M phase of the cell cycle.

MATERIALS AND METHODS
Cell Culture-KC derived from normal neonatal foreskin of single donors were purchased from Clonetics (San Diego, CA). They were cultured at low calcium concentrations in KC growth medium (Clonetics) provided by the supplier (0.15 mM). The epidermoid cell line A431 and the KC-derived cell line HaCaT (26) were cultured in either KC growth medium or in RPMI 1640 (Invitrogen) supplemented with 10% fetal bovine serum (PAA, Linz, Austria), 25 mM L-glutamine (Invitro-gen), and 1% penicillin/streptomycin (Invitrogen). 12-15 h before irradiation, KC were changed to KC basic medium (KBM; Clonetics) without growth factors. The cell lines were either changed to serum-free conditions in RPMI 1640 or to KBM. After irradiation, the culture was continued in KBM or RPMI 1640, respectively, with or without HGF/SF (20 ng/ml; R & D Systems, Minneapolis, MN). All tissue culture was performed at 37°C in 5% CO 2 and 95% air.
UV Irradiation-For irradiation experiments 1 ϫ 10 5 cells/well were seeded in 12-well plates (Costar, Cambridge, MA) and cultured overnight in KBM or RPMI 1640 without serum. Prior to irradiation, the cells were washed twice with phosphate-buffered saline, pH 7.4 (Invitrogen). UVB irradiation was carried out as described previously (27). As a light source a Mutzhas Supersun 5000-type solar simulator (Mutzhas, Munich, Germany) filtered for the emission of UVB (290 -330 nm) was used. Energy output was monitored with a IL-1700 radiometer (International Light, Newburyport, MA). Energy output was 0.7 mW/ cm 2 at a tube to target distance of 30 cm. The cells were irradiated with 8, 16, 24, and 32 mJ/cm 2 of UVB under a thin layer of phosphatebuffered saline at 25°C. For some experiments the cells were also irradiated at 4°C. For each experiment, control cells were treated identically, except for the exposure to UV light. Immediately after irradiation phosphate-buffered saline was removed, and prewarmed basic medium with or without HGF/SF was added. The cells were followed up to 5 days after irradiation.
Apoptosis Detection Assays-To quantify the extent of cell death, nonadherent cells were washed off with phosphate-buffered saline. After fixation in methanol, the remaining adherent cells were stained with methylene blue (Sigma; 0.5% in methanol). Excess methylene blue was washed out with distilled water, and the culture wells were evaluated with an inverted microscope for the presence of adherent cells. The activity of caspase-3 in cell lysates was determined according to a published protocol (28), using the synthetic caspase-3 substrate Ac-DEVD-pNA (Calbiochem, La Jolla, CA). Color development was measured on an enzyme-linked immunosorbent assay reader at 405 nm. Nucleosome release into the cytoplasm was measured with an enzymelinked immunosorbent assay according to the manufacturer's instructions (Roche Molecular Biochemicals).
Western Blot Analysis-Western blot analysis was performed as described previously (29). Briefly, 10 7 cells were lysed in 1% Nonidet P-40 lysis buffer. After quantification with the micro-␤ protein quantification kit (Pierce), the proteins were size fractionated by PAGE through an 8 -18% gradient gel (Amersham BioSciences) and transferred onto nitrocellulose membranes (Schleicher & Schü ll). Immunodetection was performed with an anti-phospho-AKT (1 g/ml, New England Biolabs, Beverly, MA), anti-phospho-MAP kinase (1 g/ml, Upstate Biotechnologies, Lake Placid, NY), and anti-c-met (1 g/ml, Upstate Biotechnologies) monoclonal antibody, followed by a horseradish peroxidase-conjugated sheep anti-mouse IgG antiserum (1:10000; Amersham BioSciences). In parallel an identical blot was reacted to an irrelevant isotype-matched monoclonal antibody as negative control. The reaction products were detected by chemiluminescence with the ECL kit (Amersham BioSciences) according to the manufacturer's instructions. Blots were quantified using the Gel-Pro Analyzer 3.1 software.
Inhibition of Kinases-For the inhibition of the PI 3-kinase and MAP kinase, KC were treated with the PI 3-kinase inhibitors wortmannin (1 M; Sigma) or LY294002 (5 M; Sigma) or with the MAP kinase inhibitor PD98059 (5 M; Biomol, Vienna, Austria) for 1 h prior to exposure to UVB. After UVB irradiation the prewarmed KBM was also supplemented with the kinase inhibitors. 15 h after irradiation, the cells were assayed for apoptotic cell death by light microscopy, nucleosome release, and caspase-3 activation.
Cell Cycle Analysis-24 and 48 h after UVB irradiation cells were incubated with bromodeoxyuridine, and 2 h later cell cycle analysis was performed using the BrdU-Flow kit (Becton Dickinson, Vienna, Austria) according to the manufacturer's instructions. Fig. 1 irradiation with 16 mJ/cm 2 induced cell death of KC (Fig. 1b) and the KC-derived cell lines A431 (Fig. 1e) and HaCaT (Fig. 1h). The addition of 20 ng/ml HGF/SF inhibited cell death and detachment of KC (Fig. 1c), whereas no inhibition was seen with A431 ( Fig. 1f) and HaCaT (Fig. 1i) cells. The induction of caspase-3activity ( Fig. 2A) and the release of nucleosomes into the cytoplasm (Fig. 2C) confirmed that both KC and KC-derived cell lines died by apoptosis. The addition of HGF/SF to UV-irradiated KC prevented nucleosome release (Fig. 2D) and caspase activation ( Fig. 2B) but had no such effect on A431 (Fig. 2, B and D) and HaCaT cells (Fig. 2, B and D). The fact that HGF/SF-treated KC remained viable after UV exposure throughout the observation period of up to 5 days proved that HGF/SF was indeed able to block rather than delay apoptosis (data not shown). Inhibition of UVB-induced apoptosis in KC could be observed at HGF/SF doses as low as 1 ng/ml (Fig. 3A) and also occurred when the factor was added 2 h after exposure to UVB. No inhibition of apoptosis was observed when HGF/SF was added 4 h after UV exposure (Fig. 3B).

UV-induced Apoptosis of KC, but Not of A431 and HaCaT Cells, Is Inhibited by HGF/SF-As shown in
The Anti-apoptotic Effect of HGF/SF Is Mediated by Signaling via PI 3-Kinase-Three different signaling pathways, i.e. the PI 3-kinase, the MAP kinase, and the STAT-3 pathways are involved in HGF/SF signaling. Blockade of the PI 3-kinase pathway by wortmannin (Fig. 4, c and d) or LY294002 (Fig. 4, e and f) completely abolished the anti-apoptotic effect of HGF/SF after UVB irradiation (Fig. 4, d and f). By contrast, the MAP kinase inhibitor PD98059 had no effect on the prevention of UVB-induced cell death (Fig. 4, g and h). As expected from these results, wortmannin and LY294002 blocked the release of nucleosomes into the cytoplasm (Fig. 5A) and the activation of caspase-3 (Fig. 5B), whereas addition of PD98059 had no such effect. Thus HGF/SF transduces the anti-apoptotic signal via the PI 3-kinase pathway.
c-met Is Functionally Active in Both KC and KC-derived Cell Lines-Because A431 and HaCaT cells were not protected from UVB-induced apoptosis by HGF/SF, we asked whether these cell lines express the receptor for HGF/SF, c-met. When we analyzed the expression of c-met in KC and A431 and HaCaT cells, proteins of identical sizes, i.e. 190 and 145 kDa under nonreducing conditions and 145 and 95 kDa under reducing conditions, were detected in primary KC and both cell lines (Fig. 6). Sequence analysis of the c-met receptor revealed no differences in the sequence of the multifunctional docking site (YV(H/N)V 16 ) between KC and KC cell lines (data not shown). To test whether c-met was functional in the KC-derived cell lines, we analyzed the phosphorylation status of AKT and MAP kinase in these cells and KC after HGF/SF stimulation. We found a strong induction of phosphorylation of both kinases in HaCaT (Fig. 7, bottom left panel) and A431 (Fig. 7, bottom right  panel) cells as well as in KC (Fig. 7, top panel). than in HaCaT and A431, respectively. For MAP kinase phosphorylation the maximal induction was 0.8 and 4.6 times stronger in KC than in HaCaT and A431, respectively. Thus activation of c-met by HGF/SF leads to phosphorylation of its target kinases in all three cell types, suggesting that the signaling pathway is intact in both KC and KC-derived cell lines. However, the fact that the maximal levels of induction of phospho-rylation were considerably stronger in KC (51-fold) as compared with both cell lines (7.4-and 4.4-fold) indicates differences in signal transduction between the different cell types.
Keratinocytes Protected by HGF/SF against UV-induced Apoptosis Enter Irreversible Cell Cycle Arrest-When after irradiation KC were cultured either in the absence or presence of HGF/ SF, we observed that in both cases the surviving cells did not proliferate but remained viable for at least up to 5 days. By cell cycle analysis after 24 h, we found a normal distribution of the different stages of the cell cycle in nonirradiated untreated cells (Fig. 8A) as well as in non-irradiated HGF/SF-treated cells (Fig.  8B). Irradiation of cells with UVB (16 mJ/cm 2 ) led to a complete loss of the S phase in the surviving cells, with a simultaneous increase in the number of cells in the G 2 /M phase (Fig. 8C). The addition of HGF/SF to UVB-irradiated KC inhibited apoptosis but did not lead to entry into S phase (Fig. 8D). Virtually the same results were obtained when cells were analyzed after 48 h (data not shown). When analyzing untreated A431 cells, we found that as compared with KC, a higher percentage of cells were in the S phase, which corresponds to their higher proliferative capacity (Fig. 8A). The addition of HGF/SF did not change the distribution of the different stages of cell cycle (Fig. 8B). In contrast to KC, irradiation of A431 cells with UVB did not lead to cell cycle arrest in the G 2 /M phase. Although most of the cells died after exposure to UVB, the surviving cells were distributed all over the different stages of the cell cycle (Fig. 8C). The addition of HGF/SF neither had an impact on A431 survival (see above) nor influenced the cell cycle distribution of surviving cells (Fig. 8D). DISCUSSION Apoptosis is the consequence of a genetically determined cell death program that can be initiated by a number of stimuli such as growth factor withdrawal, signaling through apoptotic receptors, or cell damaging stress (for review see Ref. 30). Induction of apoptosis by DNA damaging agents plays a central role in the elimination of genetically altered cells, contributing to the inhibition of tumor development (23). In the skin UV irradiation is the most relevant DNA damaging stimulus and represents the major risk factor for the development of epithelial skin tumors (22). Whereas mild UV-induced damage induces DNA repair, severe UV exposure leads to irrepairable DNA damage resulting in KC apoptosis and the formation of sunburn cells (23,24).
In the present study, we demonstrate that UVB-induced apoptosis of human KC in primary culture is completely inhibited by HGF/SF in the absence of other growth factors. In contrast to primary KC, identical treatment had no effect on the survival of the two autonomously growing KC-derived cell lines HaCaT and A431. In our experiments, we identified the PI 3-kinase/AKT pathway as the one responsible for conferring UV resistance to KC, i.e. both wortmannin and LY294002 completely abolished the anti-apoptotic effect of HGF/SF. This part of our data complements recent data by others who showed that HGF/SF confers protection against apoptosis of a renal epithelial cell line (19), of NIH 3T3 cells (21), and of hepatocytes (20) via the PI 3-kinase pathway. Blockade of the MAP kinase pathway by PD98059 had no effect on KC survival, which was surprising because activation of this pathway by insulin-like growth factor-1 has been shown to protect KC against UV-induced apoptosis (31).
Our finding that HGF/SF, although protecting KC, could not prevent apoptosis of A431 and HaCaT cells was unexpected because Tang et al. (32) have recently reported that activating the PI 3-kinase pathway via ␤ 4 integrin protects A431 cells from apoptosis. Our data exclude the possibility that the lack of protection was due to an impaired PI 3-kinase signaling, because we demonstrated that like in primary KC, activation via c-met led to AKT phosphorylation in A431 and HaCaT cells. In the light of the data of Tang et al., namely that the PI 3-kinase pathway is functional in prevention of apoptosis in A431 cells, our results suggest that in these cells, either pathways downstream of AKT cannot be activated efficiently via c-met or its activation does not lead to protection from UV-mediated apoptosis. As to the first possibility, although we could demonstrate that the MAP kinase signaling and although the activation of AKT via c-met is intact in all three cell types analyzed, we cannot rule out at present that the signaling downstream of AKT is disturbed. Phosphorylation of AKT leads to the activation and inactivation of several factors that have been shown to be involved in the induction or prevention of apoptosis. For example GSK-3, a ubiquitously expressed serine/threonine protein kinase (33); FKHR, a member of the Forkhead family of transcription factors (33,34); and Bad, an apoptotic member of the Bcl-2 family (35), are molecules that become phosphorylated by activated AKT. If any or a combination of these pathways is disturbed in the KC-derived cell lines, then no signal transduction would take place despite the presence of a functional receptor. Our finding that the amount of phosphorylated AKT was considerably lower in the cell lines than in KC suggests that in the cell lines activation of the downstream pathway might be less effective than in KC. As to the second possibility, it is tempting to speculate that different triggers of apoptosis, i.e. UV-induced DNA damage, as opposed to lack of membrane signaling via integrins might account for differing protection by HGF/SF in these cells. However, further studies will be necessary to test this hypothesis.
Furthermore, with regard to the observed differences between primary cells and autonomously growing cell lines, different pathways of cell death signaling might be the cause for the distinct anti-apoptotic effects of HGF/SF. Both HaCaT and A431 cells contain mutations in the p53 gene, which plays a central role in UV-induced apoptosis (36,37). Therefore, it is highly probable that UV-induced DNA damage leads to apoptosis via p53-dependent and -independent pathways in primary cells and cell lines, respectively. That indeed DNA damage and not UV-induced membrane effects such as the clustering of CD95 (38) was the primary responsible apoptotic trigger in our experiments is strongly suggested by our finding that UV-induced apoptosis occurred after irradiation of cells at 25°C as well as at 4°C (data not shown). Irradiation at 4°C blocks UVB-induced CD95 clustering but has no effect on UVmediated DNA damage (38).
As to the biological implications of our findings, it is important to keep in mind that in the skin dermal fibroblasts are a powerful source for HGF/SF (39). HGF/SF production is induced in these cells after stimulation with phorbol ester (40), interleukin-1 (41), and tumor necrosis factor-␣ (42). Because production the latter two factors by KC is induced by UV irradiation (43,44), their secretion would result in the triggering of HGF/SF production by fibroblasts and the establishment of a paracrine loop between epidermal and dermal symbionts, allowing the survival of KC after UV injury.
UV-induced apoptosis is thought to contribute to skin homeostasis by preferentially eliminating KC with irreparable DNA damage, i.e. potentially precancerous cells, and has been referred to as "cellular proofreading" (25), in analogy to proofreading and excision of incorrect base pairs during DNA synthesis. Indeed, mice that are deficient in p53 and therefore have a decreased capacity of DNA repair show a decreased tendency of sunburn cell formation and are more susceptible to the development of UV-induced tumors (45). For this reason UV-induced KC apoptosis by itself cannot merely be regarded as a harmful event but is thought to help reduce or prevent skin carcinogenesis (25,45). Protection of KC by the inhibition of apoptosis might therefore be a double-edged sword potentially favoring the survival of cells with DNA damage. A possibility to circumvent such a harmful scenario would be that HGF/SFprotected KC enter a postmitotic state. We could show in our experiments that KC protected by HGF/SF do not further proliferate in vitro. If this holds true in vivo, these cells would be removed from the replicating population, thereby reducing the risk of transformation while keeping them alive and thus maintaining the integrity of the skin barrier. With regard to the differences observed between cells in primary culture and cell lines bearing p53 mutations, it is tempting to speculate that HGF/SF might preferentially protect nontransformed KC in vivo while leaving transformed cells to be eliminated by UV irradiation, implying a possible beneficial effect of strong UV irradiation in the removal of precancerous epithelial skin lesions.