Administration of the antimitotic chemotherapeutic taxol is known to cause accumulation of the mitotic kinase Aurora-A (Aur-A). Here, we report that Aur-A phosphorylates S203 of the Fas associated with death domain protein (FADD) in response to taxol treatment. In addition, polo-like kinase 1 (Plk1) failed to phosphorylate the Aur-A–unphosphorylatable FADD substitution mutant S203A, indicating that phosphorylation of S203 by Aur-A serves to prime FADD for Plk1-mediated phosphorylation at S194. The double-phosphorylation-mimicking mutant form of FADD, FADD-S194D/S203D (FADD-DD), recruited caspase-8, activating the caspase-dependent cell death pathway. FADD-DD also dissociated the cell death protein RIP1 from FADD, resulting in activation of RIP1 and triggering of caspase-independent cell death. Consistent with its death-promoting potential, FADD-DD showed robust tumor suppressor activity. However, single-phosphorylation-mimicking mutant forms of FADD, FADD-S194D/S203A (FADD-DA) and FADD-S194A/S203D (FADD-AD), were incapable of carrying out such functions, indicating that double phosphorylation of FADD is critical for the execution of cell death and tumor suppression. Collectively, our data show the existence of cooperative actions between Aur-A and Plk1 mitotic kinases in response to taxol, providing a molecular explanation for the action mechanism of taxol. Cancer Res; 71(23); 7207–15. ©2011 AACR.
Antimitotic drugs comprise several classes of molecules that kill cancer cells via different cell death pathways (1). Antimitotic drugs such as taxol bind to microtubules and cause kinetic suppression of microtubule dynamics (2). The consequent cell-cycle arrest at mitotic phases by taxol leads to cell death. Although the mechanism by which taxol induces cell death may differ depending on cell type and taxol concentration, both caspase-independent and caspase-dependent pathways are involved (3–7). However, the molecular mechanism leading from cell-cycle arrest to cell death is not well understood.
Polo-like kinase 1 (Plk1) and Aurora-A kinase (Aur-A) are the most extensively studied mitotic kinases (8, 9). Plk1 has 2 functional domains: a kinase domain and polo-box domain (PBD). The PBD mediates recruitment of Plk1 to substrates that have been phosphorylated a priori by Cdk1, which serves as a priming kinase (8). Aur-A has a kinase domain in the C-terminus and an A-box in the N-terminus, which regulates the ubiquitin-dependent degradation of Aur-A (9). Plk1 and Aur-A are overexpressed in G2-M phases and coregulate multiple mitotic processes (10). Moreover, enhanced kinase activity associated with Plk1 and Aur-A overexpression confers a proliferative advantage in some tumors (11). Therefore, inhibitors of theses kinases are currently being evaluated for therapeutic use (11–16). However, the feasibility of using these kinases as anticancer drug targets is a matter of controversy. Notably, the incidence of spontaneous tumor formation is increased in mice with compromised expression of Plk1 and Aur-A, implying that these kinases possess physiologically relevant tumor suppressor potential (16–18).
FADD (Fas-associated protein with death domain) is known to interact with Fas, TNF-related apoptosis-inducing ligand (TRAIL), and TNF-1, leading to the activation of caspase-8 and subsequent cell death (19–21). Two essential domains of FADD, the death effector domain (DED) and the death domain, are involved in this process, regulating apoptosis and necrosis, respectively (22). FADD is regulated by phosphorylation, and phosphorylated FADD carries out various functions (23–25). Phosphorylation of FADD at S194, mediated by casein kinase 1α and Plk1, is required for cells to exit from G2-M (25, 26). Phosphorylated FADD also has shown death-promoting potential and enhances cellular chemosensitivity to various chemotherapeutics (23–25), including taxol. Notably, S194-phosphorylated FADD (FADD-pS194) induces caspase-independent and caspase-dependent cell death, depending on the concentration of taxol (25). The association of enhanced chemosensitivity of tumors with FADD S194 phosphorylation implies that the absence of phosphorylated FADD confers a survival advantage to tumor cells and suggests a role for FADD in the pathophysiology of cancer.
Here, we provide the first demonstration that Aur-A phosphorylates S203 of FADD in response to taxol treatment and cooperates with Plk1 in the phosphorylation of S194. Moreover, Aur-A and Plk1 double-phosphorylated FADD (dP-FADD) induces cell death via activation of caspase-8 and the cell death protein RIP1 (receptor-interacting protein 1) and suppresses tumor formation. These data suggest that mitotic kinases considered to be oncogenes play a supportive role in taxol chemotherapy. As such, administration of inhibitors against these kinases should be considered with care.
Materials and Methods
Vector construction, mutagenesis, and RNA interference
pGEX 4T-1-FADD and pFLAG-FADD construction has been previously reported (25). FADD mutations were introduced into pGEX4T-1 or pFLAG vectors using a QuikChange Mutagenesis Kit (Stratagene) with appropriate mutagenesis primers. The target sequence 5′-ATGCCCTGTCTTACTGTCA-3′ was used to develop specific siRNA against human Aur-A.
Cell line, treatments, and transfections
All cell lines were cultured in medium containing 10% FBS (Invitrogen) and 0.1% penicillin/streptomycin (Invitrogen) at 37°C in a humidified 5% CO2. HeLa (human cervical adenocarcinoma cell line), L132 (normal human lung epithelial cell line), AGS (human gastric adenocarcinoma cell line), and HCT15 (human colorectal adenocarcinoma cell line) were purchased from the Korean Cell Line Bank. HeLa and L132 were cultured in Dulbecco's modified Eagle's medium (WelGENE). AGS and HCT15 were cultured in RPMI 1640 (WelGENE). CFPAC-1 (human pancreatic adenocarcinoma cell line) was obtained from the American Type Culture Collection and cultured in RPMI 1640. All of these cell lines were free of Mycoplasma before start of experiment (Lonza). Taxol and VX680 were purchased from Sigma and Selleck, respectively. siRNA was transfected using Nucleofector (Lonza) and plasmid DNAs were transfected using METAFECTENE (Biontex), as described by the manufacturers.
Antibodies against the following proteins were purchased from the indicated sources: Aur-A and FADD (Santa Cruz Biotechnology); FADD-pS194 and caspase-8 (Cell Signaling); Plk1 (Invitrogen); FLAG (Sigma); RIP1 (BD Biosciences); and glyceraldehyde-3-phosphate dehydrogenase (GAPDH; AbFrontier). An antibody against S203-phosphorylated FADD (FADD-pS203) was raised in rabbits against a synthetic peptide corresponding to residues 199–208 of FADD-pS203 and further purified by affinity chromatography (AbFrontier).
Immunoprecipitation, immunoblotting, and immunofluorescence staining
Cell lysates were incubated with different antibodies, as indicated in the text, together with A/G-Sepharose beads (Santa Cruz Biotechnology). Immunocomplexes were resolved by SDS-PAGE, and coimmunoprecipitated proteins were detected by immunoblotting using the indicated antibodies. Immunofluorescence staining was done as previously described (27).
In vitro kinase assay
Purified glutathione S-transferase (GST)-Aur-A from Escherichia coli or immunoprecipitated Aur-A from HeLa cells transfected with pFLAG-Aur-A was incubated with purified GST-FADD mutants in kinase reaction buffer (Aur-A and Plk1 kinase assays: 20 mmol/L HEPES-KOH, pH 7.5, 5 mmol/L MgCl2, 1 mmol/L dithiothreitol, 100 μmol/L ATP; RIP1 kinase assay: 20 mmol/L HEPES-KOH, pH 7.5, 1 mmol/L NaF, 1 mmol/L Na3VO4, 20 mmol/L β-glycerophosphate, 20 mmol/L MgCl2, 20 mmol/L MnCl2, 1 mmol/L EDTA, 2 mmol/L dithiothreitol, 300 μmol/L ATP) with 1 μCi of [γ-32P] ATP at 30°C for 30 minutes. After the reactions were complete, mixtures were resolved by SDS-PAGE. The gels were stained with Ponceau S, dried, and subjected to autoradiography.
Cell death assay
Thirty-six hours after transfection, cells were harvested and used for detection of caspase-8 activity; 2.5 mmol/L IETD-AMC (Peptron) was used as a substrate. Activity was determined by measuring relative fluorescence intensity using a spectrofluorometer (PerkinElmer).
HeLa cells transfected with FADD mutants were used for colony formation and xenograft assays. These assays were done as previously described (25).
FADD interacts with Aur-A
It is known that FADD expression increases in cancer cells and tissues treated with anticancer agents (23–25, 28), suggesting that FADD might participate in the action of these agents. To elucidate the function of FADD in this context, we sought to identify FADD-interacting proteins in taxol-treated HeLa cells by incubating full-length GST-FADD with cellular proteins extracted from taxol-treated HeLa cells. Immunocomplexes with GST-FADD were analyzed by mass spectrometry (data not shown). Subsequent matrix-assisted laser desorption ionization-time of flight analysis identified several FADD-interacting molecules, including Aur-A.
To determine whether this interaction was physiologically significant, we further analyzed the association between FADD and Aur-A in HeLa cells. Extracts from cells treated with different concentrations of taxol or left untreated were prepared and immunoprecipitated with an anti-FADD antibody followed by Western blotting with Aur-A. An endogenous FADD–Aur-A complex was clearly detected in taxol-treated HeLa cells, but not in untreated cells (Fig. 1A and B). Moreover, we detected colocalization of FADD and Aur-A in taxol-treated HeLa cells, but not in untreated cells (Fig. 1C). Concordant physical interactions between FADD and Aur-A upon taxol treatment were also observed in other cells (Supplementary Fig. S1). Taken together, our data indicated that cellular complexes between FADD and Aur-A are formed upon taxol treatment, but not under unstimulated conditions.
To determine the domain(s) responsible for FADD–Aur-A interactions, we incubated GST-FADD with in vitro–translated truncated mutants of Aur-A. As shown in Fig. 1D, a region including the kinase domain of Aur-A (amino acids 133–403) interacted with FADD, whereas the N-terminus (amino acids 1–133) of Aur-A did not. These data suggest that FADD might be a substrate of Aur-A.
Aur-A phosphorylates FADD at S203 in vitro and in cells
To determine whether FADD is an authentic substrate of Aur-A, we incubated purified GST-FADD with immunoprecipitated FLAG-Aur-A WT (wild type) or KD (kinase dead) in the presence of [γ32P]ATP. FADD was phosphorylated by Aur-A-WT, but not by Aur-A-KD (Fig. 2A). Next, we substituted alanine for serine at S194, previously reported as the site of FADD phosphorylation by Plk1, generating the FADD substitution mutant, FADD-S194A (24). FADD-S194A was phosphorylated by Aur-A (Fig. 2B), indicating that S194 is not the Aur-A phosphorylation site. We next considered the S203 residue as a potential phosphorylation site for Aur-A. Although S203 is not located within a consensus motif for Aur-A substrates (R/K-X-S/T-L/I/V), it is embedded in a consensus Plk1 motif (D/E-X-S/T-ϕ; ref. 29) that is not phosphorylated by Plk1. An alanine-substituted S203 mutant FADD (FADD-S203A) was not phosphorylated by Aur-A (Fig. 2B), indicating that S203 is the FADD residue phosphorylated by Aur-A, despite its atypical (for Aur-A) sequence context.
To confirm the phosphorylation of S203 in cells, we generated an antibody specific for S203-phosphorylated FADD (FADD-pS203). A series of control experiments were done to characterize the specificity of this antibody (Supplementary Fig. S2). Both S194 and S203 phosphorylation were eradicated in lysates treated with λ-protein phosphatase (λ-PPase), indicating that both sites are authentic phosphorylation sites in cells (Supplementary Fig. S2). Moreover, phosphorylation of S203 was inhibited in Aur-A knockdown cells generated by treatment with Aur-A specific siRNA (Fig. 2C) and was significantly suppressed by treatment with the Aur-A inhibitor VX680 (Fig. 2D), which showed specific, potent activity against Aur-A kinase in vitro (Supplementary Fig. S3). Taken together, these results indicate that Aur-A phosphorylates FADD at S203.
Phosphorylation of FADD at S194 requires a priming phosphorylation of S203
Next, we examined whether phosphorylation of S203 by Aur-A has any functional relevance to S194 phosphorylation by Plk1. Upon taxol treatment, Aur-A expression was observed to increase prior to that of Plk1 (Fig. 3A). Consistent with this observation, FADD-pS203 accumulated as early as 1 hour after taxol treatment. In contrast, FADD-pS194 was detected much later, showing that phosphorylation of S203 precedes that of S194. Interestingly, knockdown and inhibition of Aur-A not only reduced the phosphorylation of S203, it also decreased S194 phosphorylation, suggesting a functional link between these two events (Fig. 2C and D). Considering that Plk1 often requires a priming phosphorylation (30), we examined whether phosphorylation of S203 affected that of S194. Phosphorylation of S203 was not affected by the phosphorylation status of S194 (Fig. 3B, left). However, phosphorylation of S194 required prior S203 phosphorylation, indicating that phosphorylation of S203 is a prerequisite for S194 phosphorylation in FADD (Fig. 3B, right). To further confirm the requirement for S203 phosphorylation, we preincubated purified GST-FADD, with or without Aur-A in the presence of unlabeled ATP, and then incubated with or without Plk1, followed by Western blotting using antibodies specific for FADD-pS194 or FADD-pS203. Purified FADD pretreated with Aur-A was strongly phosphorylated by Plk1. However, phosphorylation at S194 was significantly attenuated in the absence of Aur-A pretreatment (Fig. 3C). Furthermore, phosphorylation of S194 was strongly enhanced in cells transfected with FADD-S203D compared with that in cells transfected with FADD-S203A (Fig. 3D). To extend these observations, we examined whether the interaction between Plk1 and FADD was dependent on the phosphorylation status of S203. In pull-down experiments, FADD-WT coprecipitated with GST-Plk1-PBD, but FADD-S203A was not, indicating that the phosphorylation status of S203 determined the interaction of these two proteins (Supplementary Fig. S4).
Next, we used computer modeling to examine whether structural changes associated with phosphorylation of S203 affected phosphorylation of S194. The partial solution structure of FADD has been solved and deposited as entry 2GF5 in a protein databank (21); however, the phosphorylation sites S194 and S203 were not defined in 2GF5. To obtain the full 3-dimensional (3D) structure of FADD, including phosphorylation sites, we adopted a hierarchical protein structure modeling approach based on secondary structure, enhanced Profile–Profile threading Alignment, and the iterative implementation of the Threading ASSEmbly Refinement program (31, 32). Using this approach, we obtained a candidate model for the full 3D structure of human FADD and then carried out molecular dynamics simulations using CHARMM force field (ref. 33; version 27.0; default parameters) interfaced with Accelrys Discovery Studio 2.5 (Fig. 4A). To identify the structural changes caused by phosphorylation of S194 and S203, we also used a molecular dynamics simulations protocol in Accelrys Discovery Studio 2.5 and defined the structural modification. As shown in Fig. 4B, our modeling study predicted that phosphorylation at S203 would induce structural modifications, extending the −OH group of S194 outward and exposing it, offering a possible mechanistic basis for facilitation of S194 phosphorylation by phosphorylation of S203. In contrast, phosphorylation at S194 predicted no modification of S203, as shown in Supplementary Fig. S5. Collectively, these data showed that FADD-S203 must be phosphorylated for the subsequent phosphorylation of S194.
dP-FADD is required for the induction of cell death
The requirement for FADD phosphorylation at S194 for cell death has been previously described (23–25). Here, we studied whether the newly identified relationship between phosphorylation of FADD at S203 and phosphorylation at S194 is functionally relevant in the context of cell death. To address this, we transfected HeLa cells with various FADD mutants and examined downstream events involved in the cell death pathway. First, because FADD can induce cell death via recruitment of caspase-8, we examined the interaction of FADD mutants S194A/S203A (FADD-AA), S194D/S203D (FADD-DD), S194D/S203A (FADD-DA), and S194A/S203D (FADD-AD) with caspase-8. As shown in Fig. 5A, the FADD-DD mutant mimicking dP-FADD at S194 and S203 interacted strongly with caspase-8, whereas the phosphorylation-defective FADD-AA mutant and FADD-DA and FADD-AD mutants mimicking singly phosphorylated FADD (sP-FADD) did not. Second, we examined whether dP-FADD affected caspase-8–mediated DEF (death effector filament) formation (34). FLAG-tagged FADD mutants were transfected into HeLa cells with pEGFP–caspase-8 DEDs. FADD-DD strongly induced caspase-8–mediated DEF formation, whereas unphosphorylated FADD or mutants mimicking sP-FADD did not (Supplementary Fig. S6). Furthermore, FADD-DD induced cell death via activation of caspase-8 and caspase-3 (Fig. 5B; Supplementary Fig. S7), implying that dP-FADD induces caspase-mediated cell death. FADD also plays an important role in the necrotic branch of cell death (22). To determine whether RIP1 is activated by dP-FADD, we transfected HeLa cells with FLAG-tagged FADD mutants. Forty-eight hours after transfection, cell lysates were immunoprecipitated with an anti-FLAG or anti-RIP1 antibody, followed by Western blotting with an anti-RIP1 antibody (Fig. 5C) or kinase assays (Fig. 5D). As shown in Fig. 5C and D, FADD-DD dissociated RIP1 from FADD and induced activation of RIP1, whereas mutants mimicking unphosphorylated FADD (FADD-AA) or sP-FADD (FADD-DA and FADD-AD) did not. Such requirement of double phosphorylation for cell death was confirmed in various cells including HeLa (Supplementary Fig. S8). Collectively, these data showed that double phosphorylation is a critical phenomenon for both apoptosis and necrosis.
Phosphorylation of FADD at S194 increases the sensitivity of tumor cells to taxol (24). Therefore, we were curious whether FADD S203 also confers to the chemosensitivity to taxol. Sensitivity of HeLa cells to taxol was decreased when treated with BI2536 (Plk1 inhibitor) and VX680 (Aur-A inhibitor), implying that phosphorylations of FADD at S194 and S203 by Plk1 and Aur-A are important in taxol-induced cell death (Supplementary Fig. S9). Next, we examined whether taxol-induced cell death is mediated via activation of caspase-8 and RIP1. When we treated HeLa cells with necrostatin-1 (RIP1 inhibitor) and z-IETD-fmk (caspase-8 inhibitor), taxol-induced cell death was completely abolished, indicating that taxol-induced cell death requires activation of caspase-8 and RIP1 (Supplementary Fig. S10).
FADD serves as an adaptor in the extrinsic cell death pathway (19–21). To determine whether phosphorylation of FADD is involved in the receptor-mediated cell death, FADD-negative Jurkat cells (FADDdef) were transfected with FADD-DD or FADD-AA and subsequently treated with TRAIL. Reconstituted cells were equally sensitive to TRAIL-mediated cell death (Supplementary Fig. S11). Moreover, treatment of VX680 did not block the TRAIL-induce cell death (Supplementary Fig. S12). Collectively, our data indicated that phosphorylation of FADD does not contribute to TRAIL-mediated cell death. These findings were consistent with the previous study reporting that phosphorylation of FADD was irrelevant in Fas-mediated cell death (35).
dP-FADD suppresses tumor formation
We previously showed that FADD-pS194 inhibits Plk1-mediated tumor formation (25). Here, we compared the tumor suppressor activities of sP-FADD and dP-FADD by expressing FLAG-tagged mutants of FADD in HeLa cells and then analyzing in vitro colony formation and in vivo tumorigenesis. In the in vitro colony-formation assay, transfection of FADD-DD significantly inhibited cell growth compared with transfection of FADD-AA, FADD-DA, FADD-AD, or empty vector (MOCK). This indicates that dP-FADD regulated cell growth more efficiently than sP-FADD and unphosphorylated FADD (Fig. 6A). Next, we analyzed the effects of FADD mutants on inhibition of tumor growth in nude mice. HeLa cells overexpressing FADD mutants (2 × 106 cells per mouse) were injected into 5-week-old nude mice; 5 weeks later, the tumors formed in these mice were dissected for measurement of tumor size (Fig. 6B, left). Tumor growth was significantly inhibited in mice injected with HeLa cells transfected with FADD-DD compared with those transfected with FADD-AA, FADD-DA, FADD-AD, or MOCK (Fig. 6B). This implied that sP-FADD is not sufficient to suppress tumor growth. Therefore, we speculated that the previously reported tumor suppressor activity of FADD-pS194, which would have included both dP-FADD and singly phosphorylated FADD-pS194 forms (i.e., not phosphorylated at S203), is actually attributable to dP-FADD (25). Moreover, our computer modeling suggested that the possibility that singly phosphorylated FADD-pS194 exists in cells would be low.
This study identified S203 as a novel phosphorylation site of FADD, in addition to the previously identified S194 site, and describes the implication of this phosphorylation. Moreover, we report that FADD is double phosphorylated by the sequential action of Aur-A and Plk1 upon taxol treatment and show that dP-FADD activates both caspase-dependent and caspase-independent pathways, leading to cell death. In addition, dP-FADD successfully suppressed tumor formation, whereas sP-FADD did not. Therefore, our data indicate that the cooperative action of the two mitotic kinases Aur-A and Plk1 is essential for endowing the death-promoting and tumor suppressor functions of FADD.
Plk1 binds to phosphoserine- or phosphothreonine-containing peptides through its PBD (36). Sequential phosphorylation of substrates, including vimentin, BubR1, Cep55, and Bora, by Cdk1 and Plk1 have been reported (37–40), where Cdk1 serves a common priming kinase function that facilitates subsequent interactions between substrates and Plk1 (8). However, Cdk1 does not phosphorylate FADD (26). In this study, we found that Aur-A serves as a priming kinase for subsequent phosphorylation of FADD by Plk1, adding FADD to the list of Plk1 substrates that require a priming kinase. Such a cooperative phosphorylation by Aur-A and Plk1 has not previously been reported. Kif2a, Bora, and p150glued are also phosphorylated by Aur-A and Plk1; however, cooperative phosphorylation has not been shown (41–44). Therefore, this study presents a novel regulatory mechanism involving the action of Aur-A and Plk1 upon their common substrate, FADD.
Phosphorylation by Aur-A confers a “gain of function” to FADD. If phosphorylation of S203 played only a priming role for subsequent phosphorylation of S194, overexpression of FADD-DA or FADD-DD should yield equivalent phenotypes. However, overexpression of FADD-DD potently killed cells and suppressed tumorigenesis in vivo, whereas overexpression of FADD-DA did not. This indicates that phosphorylation of FADD at S203 by Aur-A has other effects in addition to its role in priming FADD for subsequent action by Plk1. Structural changes in FADD induced by sequential phosphorylation events seemed to affect its interactions with other proteins. Unphosphorylated FADD and sP-FADD remained bound to RIP1, whereas dP-FADD dissociated RIP1 and recruited caspase-8. This provides a molecular explanation for the selective death potential of dP-FADD. It is well known that recruitment of caspase-8 to FADD leads the activation of caspase-8. However, the RIP1 activating mechanism coupled with its dissociation from dP-FADD has yet to be clarified. It would be interesting to examine whether interaction of RIP1 with RIP3, a RIP1 kinase, increases upon dissociation of RIP1 from dP-FADD. FADD has been considered to function as a selective molecular switch that governs apoptotic versus and necrotic outcomes (22). However, our data show that dP-FADD activates both types of cell death, implying that dP-FADD is a powerful weapon in the cell-death arsenal.
The functions of Plk1 and Aur-A are a matter of controversy. These kinases are overexpressed in various human cancers, and their high levels of expression are associated with poor prognosis (45, 46). These kinases are considered oncogenes, and many pharmaceutical companies are developing cancer drugs based on inhibition of their catalytic functions (47). However, data with the opposite implication have also been reported. Overexpression of Plk1 suppresses cell proliferation and mutations causing inactivation of Aur-A increase the risk of developing cancer (17, 18, 48). The data presented here also show that increased activities of Plk1 and Aur-A endow FADD with death-promoting and tumor suppressor functions.
In our view, administering inhibitors of Aur-A and Plk1 to cancer patients with uncompromised expression of FADD might not be appropriate because this would prevent dual phosphorylation of FADD. Instead, we think that drugs that induce accumulation of Aur-A and Plk1, such as taxol, might be of benefit to these patients. Further studies will ultimately be required to clarify whether sacrificing the death potential of dP-FADD is ultimately beneficial to such patients. Our data do suggest, however, that inhibitors of Plk1 and Aur-A might be useful for cancer patients with low-level expression of FADD. Therefore, the expression level of FADD in cancer patients might serve as a decisive parameter for determining which type of chemotherapeutics should be administered.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
The authors thank Hye-Lim Jang (Samsung Biomedical Research Institute, Samsung Medical Center, Sungkyunkwan University, School of Medicine, Seoul, Korea) for assisting in the construction of FADD mutants.
This work was supported by National Research Foundation grant funded by the Korea government (MEST; 2010-0029609).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.