006) 1958–1966www.elsevier.com/locate/cellsig

Cellular Signalling 18 (2

A role for extracellular and transmembrane domains of Sefin Sef-mediated inhibition of FGF signaling

Dmitry Kovalenko a, Xuehui Yang a, Pei-Yu Chen a,b, Robert J. Nadeau a,b,Olga Zubanova a, Kathleen Pigeon a, Robert Friesel a,b,⁎

a Center for Molecular Medicine, Maine Medical Center Research Institute, 81 Research Dr., Scarborough, Maine 04074-7205, USAb Cooperative Graduate Program, Molecular Genetics and Cell Biology, University of Maine, Orono, Maine 04469, USA

Received 1 February 2006; received in revised form 2 March 2006; accepted 3 March 2006Available online 10 March 2006

Abstract

Sef (similar expression to fgf genes) is a member of the fibroblast growth factor (FGF) synexpression group that negatively regulates FGFreceptor (FGFR) signaling in zebrafish during early embryonic development and in mammalian cell culture systems. The mechanism by which Sefexerts its inhibitory effect remains controversial. It has been reported that Sef functions either through binding to and inhibiting FGFR1 activationor by acting downstream of FGF receptors at the level of MEK/ERK kinases. In both cases, the intracellular domain of Sef was found to play arole in the inhibitory function of Sef. Here we demonstrated that both extracellular and transmembrane domains of Sef contributed to Sef-mediatednegative regulation of FGF signaling. In fact, over-expression studies in NIH3T3 cells showed that a truncated mutant of Sef, which lacks theintracellular domain (SefECTM), exerted the inhibitory activity on FGF signaling by inhibiting FGFR1 tyrosine phosphorylation and subsequentactivation of the Raf/MEK/ERK signaling cascade. We also showed that SefECTM associated with FGFR1, and inhibited FGF-induced ERKactivation in HEK293T cells. Furthermore, we demonstrated that the over-expression of SefECTM was able to inhibit the function of aconstitutively activated form of FGFR1, FGFR1-C289R, but not FGFR1-K562E. Finally, we found that SefECTM reduced cell viability whenover-expressed in human umbilical vein endothelial cells (HUVEC). These data provide additional insight into the structure–activity relationshipin the mechanism of inhibitory action of Sef on FGFR1-mediated signaling.© 2006 Elsevier Inc. All rights reserved.

Keywords: Sef; IL-17RD; FGF; FGFR; Signaling; MAPK; HUVEC; Apoptosis

1. Introduction

The fibroblast growth factor (FGF)/FGF receptor (FGFR) isan evolutionary conserved signaling system that plays a crucial

Abbreviations: Sef, similar expression to fgf genes; FGF, fibroblast growthfactor; FGFR, FGF receptor; SH, Src homology; ERK, extracellular signal-regulated kinase; MEK, mitogen-activated protein kinase/extracellular signal-regulated kinase kinase; RTK, receptor tyrosine kinase; PDGF, platelet-derivedgrowth factor; GFP, green fluorescent protein; DMEM, Dulbecco's modifiedEagle's medium; CS, calf serum; MAPK, mitogen-activated protein kinase;HUVEC, human umbilical vein endothelial cells.⁎ Corresponding author. Center for Molecular Medicine, Maine Medical

Center Research Institute, 81 Research Dr., Scarborough, Maine 04074-7205,USA. Tel.: +1 207 885 8147; fax: +1 207 885 8179.

E-mail address: friesr@mmc.org (R. Friesel).

0898-6568/$ - see front matter © 2006 Elsevier Inc. All rights reserved.doi:10.1016/j.cellsig.2006.03.001

role in many developmental and physiological processes inanimal species, including formation and remodeling of bloodvessels, wound healing, organ morphogenesis, skeletal develop-ment, and early embryogenesis (for review, see [1]). Genetargeting of FGF or FGFR gene family members in the mousedemonstrates that FGF signaling is essential for cell proliferationand survival in the preimplantation mouse embryo, as well as forcell migration during gastrulation [2,3]. Aberrant functioning ofFGF signaling as a result of activating mutations in FGFR genesis associated with several pathological conditions in humans,primarily with craniofacial and limb malformations, and sometypes of malignancies [1,4]. Twenty-four members of the FGFgene family and five members of the FGF receptor gene familyhave been identified [3,5,6]. FGFs are structurally similarpolypeptides that are expressed at various stages of development

Fig. 1. Sef regulates ERK activation induced by constitutively active forms ofFGFR1. (A) The level of endogenous Sef mRNA expression in HEK293T cellsco-transfected with FGFR1-C289R, ERK2-GFP, and either a negative controlsiRNA (Ctrl) or anti-Sef siRNA (αSef) was analyzed using RT-PCR. (B)HEK293T cells transfected with FGFR1-C289R (0.1 μg) and ERK2-GFP(0.5 μg) constructs together with either a negative control siRNA (4 μg) or anti-Sef siRNA (4 μg) were serum starved for 24 h. Cell lysates were subjected toimmunoblot analysis with anti-p-ERK1/2 (p-ERK2-GFP), anti-ERK1/2 (ERK2-GFP), and anti-FGFR1 (clone 5G11) antibodies.

1959D. Kovalenko et al. / Cellular Signalling 18 (2006) 1958–1966

in the epithelium, muscle, connective tissue and nerves, wherethey regulate cell proliferation, survival, adhesion, motility,senescence, and differentiation [3].

FGFs elicit their biological activities by binding to andactivating FGFRs, which are evolutionarily conserved, single-pass transmembrane proteins belonging to the super-family ofreceptor tyrosine kinases [7]. The binding of FGF ligands toFGFR induces receptor dimerization and autophosphorylationon specific tyrosine residues of the intracellular domain [8].Subsequent to receptor phosphorylation, the SH2 (Srchomology) domain-containing proteins, Crk and PLCγ, bindto specific phosphotyrosines of FGFR and mediate thetransmission of activating signals to a variety of cytoplasmicproteins [9–11]. An adaptor protein FRS2 binds to FGFR in aphosphotyrosine-independent manner, and is tyrosine-phos-phorylated upon activation of FGFRs [12]. Once FRS2 istyrosine-phosphorylated, it binds the SH2 domain-containingadaptor protein Grb2 [12]. Grb2 recruits SOS to the plasmamembrane where it participates in activation of a lowmolecular weight G-protein Ras [13–15]. Activated Rasbinds a cytoplasmic kinase Raf resulting in the localizationof the latter on the inner surface of the plasma membrane, andits subsequent activation [16]. Raf, in turn, phosphorylates andactivates MEK, which has dual activities and is capable ofphosphorylating serine/threonine and tyrosine residues of themitogen-activated protein kinases (MAPK) ERK1 and ERK2[17,18]. Activated ERK1/2 phosphorylates other cytoplasmicsubstrates and nuclear transcription factors, thus modulatingtheir activities [19,20].

FGF signaling is tightly regulated temporally and at multiplelevels of the signaling cascade. One mode of such regulationincludes temporal modulation of the intensity and duration ofFGF signaling by positive and negative feedback mechanisms[21]. During early embryogenesis, FGF signaling is limited byaction of feedback antagonists, such as Sef, Sprouty, SPRED(Sprouty-related EVH1-domain-containing), MKP1, and MKP3(MAPK phosphatase 1 and 3) proteins [21]. In contrast, therecently discovered protein XFLRT3 (Xenopus fibronectin-leucine-rich transmembrane protein 3) was shown to act as apositive feedback regulator of FGF signaling during earlyXenopus development [22]. Because Sef, Sprouty2, Sprouty4,and XFLRT3 have a common spatial and temporal pattern ofexpression with FGF3 and FGF8 genes during embryogenesis,they were classified as members of the FGF synexpression group[21].

Sef was originally described in zebrafish (zSef) as a putativetype I transmembrane protein that negatively regulates FGFsignaling during early embryonic development [23,24]. It hasbeen reported that zSef associates with FGFR1 and FGFR2 inco-immunoprecipitation assays [23]. Subsequent cloning, mu-tagenesis, and over-expression studies of mouse and human Sef(mSef and hSef) revealed that they are similarly able toassociate with FGFR1 and FGFR2, and inhibit FGF signaling indifferent cell culture systems [25,26]. Those studies indicated arole for the intracellular domain of Sef in its inhibitory function.Nevertheless, the precise mechanism by which Sef elicits itsinhibitory effect remains controversial. It has been reported that

Sef acts at or downstream of MEK [24,27,28]. Alternatively, aswe have recently demonstrated, mSef may inhibit FGF signalingin NIH3T3 cells by binding to and inhibiting FGFR1 tyrosinephosphorylation and subsequent Raf/MEK/ERK pathwayactivation [25]. We have also reported that the intracellulardomain of mSef is primarily responsible for the mSef–FGFR1interaction [25]. In the present study, we demonstrated that over-expression of a truncated mutant of mSef with a deletedintracellular domain (mSefECTM) inhibited FGF-inducedtyrosine phosphorylation of FGFR1, and subsequent activationof the Raf/MEK/ERK signaling cascade in NIH3T3 cells, aswell as downregulated FGFR-dependent ERK activation inHEK293T cells. Furthermore, we demonstrated that over-expression of mSefECTM reduced the viability of HUVEC,whose survival is strictly dependent upon the steady-state levelof FGF signaling activation. We also showed that mSefECTMphysically associated with FGFR1. Thus, our present dataclarify the domain requirements for Sef-mediated inhibition ofFGFR-dependent signaling.

2. Materials and methods

2.1. Materials

2.1.1. Expression vectorsPlasmids encoding mSefFL, mSefECTM, mSefECpTM (previously denoted

as mSefEC), and mSefICTM (previously denoted as mSefIC), were describedpreviously (see Fig. 1A in [25]). The deletion mutant mSefICpTM wasconstructed by PCR using primers designed to amplify the fragment of mSef(amino acids 321–738) (GenBank™ accession no. AF459444), and subclonedinto the pDisplay plasmid (Invitrogen) to add the PDGFR transmembranedomain, followed by subsequent subcloning into pcDNA3.1/V5-His TOPOvector (Invitrogen). The sequences of mSef mutants were confirmed using anABI 310 automated DNA sequencer. Plasmids encoding FGFRs were describedpreviously [29]. For the preparation of mSefFL-, mSefICpTM-, mSefECTM- or

1960 D. Kovalenko et al. / Cellular Signalling 18 (2006) 1958–1966

GFP- (hrGFP, Stratagene) expressing adenoviruses, a Cre-lox-based system wasemployed [30] (gift from P. Robbins, University of Pittsburgh).

2.1.2. AntibodiesAntibodies used in this study were rabbit anti-FGFR FB817 [33]; anti-

FGFR1 monoclonal antibody 5G11 [34], anti-V5 (Invitrogen); anti-p-ERK1/2(Sigma, M8159); anti-phosphotyrosine (pY) 4G10, anti-ERK1/2 (UpstateBiotechnology, Inc.), and anti-cleaved caspase 3 (Cell Signaling).

2.2. Cell cultures, treatments and transfection

2.2.1. Cell lines, transfections, adenoviral transduction and MTT assayHEK293T cells (ATCC) and NIH3T3 cells (ATCC) were maintained in

Dulbecco's modified Eagle's medium (DMEM) containing 10% FBS(HyClone) or 10% CS (HyClone), respectively. HUVEC (passages 3–7)were propagated in M199 medium (Invitrogen) containing 10% FBS(HyClone) and 10 ng/ml FGF2 (R&D Systems). For transient transfectionexperiments, HEK293T cells were plated at 40–50% confluence on 6 cmdishes, and grown overnight in growth medium. Transfections were performedwith the Genejuice reagent according to the manufacturer's protocol(Novagen). Six hours after transfection, the cells were trypsinized, replatedon fibronectin coated 6-well plates, serum-starved overnight (∼12 h) andstimulated with FGF2 (10 ng/ml) for 10 min. NIH3T3 cells were transducedwith AdGFP, AdmSefFL, AdmSefICpTM or AdmSefECTM (104 viralparticles/cell), and complexed with poly-lysine (Sigma), as described [31].Six hours after transduction, the cells were trypsinized, replated on fibronectincoated 10 cm dishes, serum starved overnight (∼12 h), and stimulated withFGF2 or PDGFBB (R&D Systems) (10 ng/ml of each) for 10 min. HUVECwere transduced by incubating cells with the indicated adenoviruses in growthmedium (1000 viral particles/cell). MTT assays was performed as previouslydescribed [32].

2.2.2. siRNA transfection and RT-PCRFor siRNA transfection experiments HEK293T cells were plated at 40–50%

confluence on fibronectin-coated six well plates. Co-transfections of theindicated amounts of plasmid DNA and siRNAwere performed with the siPortAmine Transfection Agent (Ambion) (20 μl/well) according to the manufac-turer's protocol. Anti-Sef siRNA (sense: 5′-GGCUUGGUUUUAAAUGAUGtt-3′ and antisense: 5′-CAUCAUUUAAAACCAAGCCtg-3′) and control siRNA(Negative Control siRNA#1) were obtained from Ambion (Austin, TX). TotalRNA was isolated from transfected HEK293T cells using RNeasy Mini Kit(Qiagen) as recommended by the manufacturer. RNA concentrations were mea-sured by spectrophotometry at 260 nm. For each sample 4 μg of total RNAwasused to synthesize cDNA using Cloned AMV First-Strand cDNA Synthesis Kit(Invitrogen). This cDNA reaction mixture (20 μl) was diluted with 180 μl ofDNAase/RNAase-free water and 5 μl of the cDNA solution was used for genespecific PCR performed with Invitrogen's PCR SuperMix. The PCR primersused were: Sef sense: 5′-GCCTGTGACCTG TTGTTACA-3′, Sef antisense: 5′-GAACTGGGACTCGTGGATCT-3′; β-actin sense: 5′-TGTTACCAACTGG-GACGACA-3′, β-actin antisense: 5′-CTCTCAGCTGTGGTGGTGAA-3′. β-actin was amplified as a housekeeping gene to normalize expression. Samples notundergoing reverse transcription were run in parallel as controls. PCR wasperformedwith each cycle consisting of 94 °C for 1min, 50 °C for 30 s, and 72 °Cfor 1 min, followed by a final extension step at 72 °C for 7 min. PCR cyclenumbers were kept low to perform semi-quantitative PCR (actin: 20 cycles, Sef:35 cycles). PCR products were resolved on 1% ethidium bromide-stainedagarose gels, and visualized by ultraviolet transillumination.

2.2.3. Immunoprecipitation and immunoblottingCells were lysed in HNTG buffer (20 mM HEPES, pH 7.4, 150 mM NaCl,

10% glycerol, 1% Triton X-100, 1.5 mM MgCl, 1.0 mM EGTA, 0.1 mMNaVO4) as described [33]. Cell lysates were subjected to immunoblot analysis,either directly or after immunoprecipitation with the indicated antibodies.Immune complexes were captured on protein A/G-agarose (Santa Cruz), washedextensively in HNTG buffer, eluted with 2×SDS sample buffer, separated on8% SDS-PAGE, and transferred to Hybond-C (Amersham), followed byimmunoblotting as described [33]. Bound antibodies were visualized bychemiluminescence (ECL reagent, Amersham). The intensity of the bands was

determined by scanning with a Canon scanner (model N1220), and analysis withNIH image software.

2.2.4. Cross-linkingTransfected HEK293Tcells were incubated overnight in serum-free medium

(DMEM). Cells were stimulated with or without FGF2 (10 ng/ml) for 10 min,washed with phosphate-buffered saline (PBS), and incubated in PBS containing10 mM of the homobifunctional cross-linker DTSSP (3,3-Dithiobis[sulfosucci-nimidyl propionate]) (Pierce, Rockford, IL) for 10 min at 37 °C. Afterwards,cells were washed three times with PBS/10 mMTris–HCl (pH 7.5) solution, andlysed in HNTG buffer. Cell lysates were either immunoblotted directly with anti-V5 (Sef) or immunoprecipitated with anti-FGFR antibodies (FB817), followedby immunoblotting with anti-V5 or anti-FGFR (5G11) antibodies.

3. Results

3.1. Sef regulates ERK activation induced by a constitutivelyactive FGFR1

It has recently been established that the over-expression ofboth human and mouse Sef results in inhibition of FGFR-dependent ERK (p42/44 MAPK) activation in several cell types[25–27]. In order to determine whether endogenous Sef is ableto regulate FGFR1-mediated ERK activation, we performedsilencing of Sef mRNA expression in HEK293T cells whichalso were transiently transfected with plasmids encoding aconstitutively active form of FGFR1, FGFR1-C289R [29], andERK2-GFP (Fig. 1A). Quiescent cells were lysed and celllysates were subjected to immunoblotting with antibodies thatspecifically recognize the phosphorylated (activated) form ofERK1/2. Immunoblot analysis revealed that siRNA silencing ofendogenous Sef expression in HEK293T cells led to an increasein ERK phosphorylation induced by FGFR1-C289R (Fig. 1B).

3.2. Contribution of different domains of Sef to mSef-mediatedinhibition of FGF signaling

We have previously reported that full-length mouse Sef(mSefFL), as well as its truncated mutant (mSefICTM) thatcontains intracellular and transmembrane portions of Sef, co-immunoprecipitated with FGFR1 and mediated inhibition ofFGF-induced ERK phosphorylation; however, the inhibitoryactivity of mSefICTM was significantly reduced whencompared with mSefFL [25]. Those results indicated that apart of the inhibitory activity of full-length Sef may be attributedto its extracellular domain. In an effort to determine whether theextracellular domain-containing Sef mutants (mSefECTM andmSefECpTM) are able to inhibit FGF-induced ERK activation,we transfected HEK293T cells with either full-length mSef orSef deletion mutants expressing constructs (Fig. 2A) togetherwith a plasmid encoding ERK2-GFP. Eighteen hours aftertransfection, quiescent cells were stimulated with FGF2 for10 min, and activation of ERK2 was determined by immuno-blotting with antibodies that recognize the phosphorylated formof ERK1/2. The over-expression of mSefFL, mSefICTM,mSefICpTM, and mSefECTM resulted in the reduction ofERK2 activation, while over-expression of mSefECpTM had noeffect on FGF2-induced ERK2 phosphorylation (Fig. 2B).Remarkably, the replacement of the Sef transmembrane domain

Fig. 3. mSefECTM inhibits FGF-induced FGFR1 activation. (A) NIH3T3 cellswere transduced with adenoviral vectors expressing the indicated proteinsfollowed by serum starvation for 12 h. Quiescent cells (Q) were stimulated withFGF2 (10 ng/ml) for 10 min, and cell lysates were subjected to immunopre-cipitation with anti-FGFR antibodies (FB817), followed by immunoblottingwith anti-phosphotyrosine antibodies (anti-pY, clone 4G10). Blots were strippedand re-probed with anti-FGFR antibodies (FB817). (B) Relative intensities ofFGFR1 tyrosine phosphorylation were determined by scanning and subsequentquantification of the immunoblots by utilizing the NIH Image software. Over-expression of mSefFL, mSefICpTM, and mSefECTM resulted in 72±5%, 28±7%, and 64±18% (mean±S.D.) decrease in FGFR1 tyrosine phosphorylation,respectively, as calculated on a total level of FGFR1 adjusted basis (n=3).

Fig. 2. Over-expression of mSef or its truncated mutants in HEK293T andNIH3T3 cells inhibits FGF-induced ERK activation. (A) mSef expressionconstructs. Black square indicates the mSef signal sequence, black triangle —the Igκ signal sequence substitution, dark gray— the transmembrane domain ofmSef (TM), and light gray — the PDGFR transmembrane domain (pTM)substitution. All constructs have a C-terminal V5/His tag (not shown). (B) mSef(5 μg) or mSef deletion constructs with ERK2-GFP plasmid (0.5 μg) weretransiently transfected in HEK293T cells. Quiescent (Q) cells were stimulatedwith FGF2 (10 ng/ml) for 10min and cell lysates were immunoblotted with anti-p-ERK1/2 (p-ERK2-GFP), anti-ERK1/2 (ERK2-GFP), and anti-V5 (Sef)antibodies. (C) Quiescent NIH3T3 cells transduced with indicated adenoviralvectors were stimulated with FGF2 (10 ng/ml) or PDGF-BB (10 ng/ml) for10 min, followed by subjecting cell lysates to immunoblotting with antibodiesagainst p-Raf1, p-MEK1/2, and p-ERK1/2. The expression of mSef proteins andthe equivalence of loading were shown by re-blotting with anti-V5 and β-actinantibodies, respectively.

1961D. Kovalenko et al. / Cellular Signalling 18 (2006) 1958–1966

in mSefECTM with the PDGFR transmembrane domain(mSefECpTM) had a dramatic effect on mSefECTM inhibitoryactivity, while a similar replacement in mSefICTM did notreduce its ability to downregulate FGF-induced ERK activation(Fig. 2B). Similar results were obtained when mSefFL and itstruncated mutants were over-expressed in HEK293T cells witha wild type FGFR1 (data not shown). These data indicated thatwhile the presence of the transmembrane domain of Sef wasrequired for mSefECTM inhibitory activity, it did not contributeto mSefICTM-mediated FGF signaling inhibition.

In an attempt to determine the point in the Raf/MEK/ERKpathway where the inhibition of FGF signaling occurs, wetransduced NIH3T3 cells with adenoviral vectors expressingmSefFL, mSefICpTM or mSefECTM. Eighteen hours aftertransduction, quiescent cells were stimulated with FGF2 orPDGFBB for 10 min, and immunoblotting of cell lysates wasperformed by using antibodies that recognize the phosphory-

lated (activated) form of Raf-1 (Ser388), MEK1/2 (Ser217/Ser221), and ERK1/2 (Thr202/Tyr204). Immunoblot analysisrevealed that the over-expression of mSefFL, mSefICpTM ormSefECTM proteins significantly reduced activation of thesekinases (Fig. 2C). These data indicated that the site of action ofmSefECTM, as well as mSefICpTM, is likely to be upstream ofRaf-1, and are consistent with our previously published resultsthat mSefFL inhibits FGF signaling upstream of Ras [25]. Thelack of the ability of either mSefFL or its truncated mutants toinhibit activation of the Raf/MEK/ERK cascade, induced bystimulation with PDGFBB, suggests that signaling from Rasremains intact in those cells.

3.3. mSefECTM exerts its inhibitory effect by inhibiting FGFR1phosphorylation

Since, as we previously reported, mSefFL inhibits FGFsignaling through a mechanism that includes Sef binding to andinhibiting FGFR1 activation [25], we next sought to determinewhether mSefECTM affects FGF-induced FGFR1 tyrosinephosphorylation. Therefore, quiescent NIH3T3 cells transducedwith indicated recombinant adenoviral vectors were treated withor without FGF2, and the cell lysates were subjected toimmunoprecipitation with anti-FGFR antibodies, followed byimmunoblotting with anti-phosphotyrosine and anti-FGFRantibodies (Fig. 3A). The results indicate that adenovirus-mediated over-expression of both mSefFL and mSefECTMreduced FGF-induced tyrosine phosphorylation of FGFR1without affecting the overall level of FGFR expression(Fig. 3A). Interestingly, although mSefICpTM also inhibited

Fig. 4. mSefECTM associates with FGFR1. HEK293T cells were co-transfectedwith mSefECpTM or mSefECTM constructs with FGFR1 encoding plasmid.Quiescent (Q) cells were stimulated with or without FGF2 (10 ng/ml), andsubsequently cell surface proteins were cross-linked by incubating cells withDTSSP cross-linker, as described in the “Experimental Procedures”. Cell lysateswere immunoprecipitated with anti-FGFR antibodies (FB817), and immuno-precipitates were subjected to immunoblotting with anti-V5 (Sef) and anti-FGFR1 (5G11) antibodies. To demonstrate the total level of mSefECpTM ormSefECTM expression, cell lysates (CL) were immonoblotted with anti-V5antibodies.

Fig. 5. Over-expression of mSef or its subdomains in HUVEC decreases cell viabilitmSef truncated mutants were taken at the indicated times after transduction. (B) Thevectors was determined by using the MTTassay. Data are representative of three expewere prepared 48 h after transduction, and subjected to immunoblotting with antibodMAPK. The level of Sef expression was visualized by using anti-V5 antibody.

1962 D. Kovalenko et al. / Cellular Signalling 18 (2006) 1958–1966

FGFR1 tyrosine phosphorylation, the total level of the receptorwas also reduced (Fig. 3A). Densitometric analysis ofimmunoblots of three independent experiments with subsequentadjustment of intensities of the tyrosine phosphorylated bandsto the total levels of FGFR1 indicated that the over-expressionof mSefFL, mSefICpTM, and mSefECTM mediated a 72%,28%, and 64% decrease in a ligand-induced FGFR1 tyrosinephosphorylation, respectively (Fig. 3B).

3.4. mSefECTM interacts with FGFR1 and its transmembranedomain is required for this interaction

As we have previously reported, the intracellular portion ofmSef primarily mediates mSef–FGFR1 binding, while themSefECTM–FGFR1 complex is not efficiently recovered by astandard co-immunoprecipitation technique [25]. We surmisedthat the detergent extraction and solubilization of transmem-brane proteins, such as mSefECTM and FGFR1, could interfere

y. (A) Phase-contrast photomicrographs of HUVEC expressing GFP, mSefFL orcell viability in populations of HUVEC transduced with the indicated adenoviralriments. (C) Cell lysates of HUVEC expressing the indicated Sef protein variantsies against cleaved caspase-3 and phosphorylated forms of ERK1/2, JNK or p38

Fig. 6. mSefECTM specifically inhibits FGF-induced ERK activation inHUVEC. (A) HUVEC were grown for 24 h in the growth medium containing10% serum with of without FGF2 (10 ng/ml). Subsequent immunoblot analysiswas performed by using anti-p-MEK, anti-p-ERK1/2, and anti-ERK1/2antibodies. (B) Cell lysates of HUVEC transduced with AdGFP, AdECTM orAddnFGFR for 24 h were subjected to immunoblotting with antibodies againstp-MEK1/2, p-ERK1/2 or ERK1/2. (C) HUVEC were transduced with ade-noviral vectors expressing GFP or mSefECTM proteins, followed by serumstarvation for 6 h. Quiescent cells (Q) were stimulated with FGF2 (10 ng/ml) orserum (FBS, 10%) for 10 min, and cell lysates were subjected to immunoblotanalysis with anti-p-ERK1/2, ERK1/2, and V5 (Sef) antibodies.

1963D. Kovalenko et al. / Cellular Signalling 18 (2006) 1958–1966

with a possible hydrophobic interaction which may occurbetween transmembrane domains of proteins within a plasmamembrane compartment. In order to overcome this technicallimitation, we performed cross-linking of cell surface proteinsprior to a co-immunoprecipitation assay. HEK293T cells weretransfected with FGFR1-expressing plasmid together withmSefECTM or mSefECpTM constructs. Eighteen hours aftertransfection, quiescent cells were treated with or without FGF2for 10 min. Cross-linking reactions were performed utilizing amembrane-impermeable cross-linker, DTSSP (Pierce). Sefproteins that were associated with FGFR1 in co-immunopreci-pitated complexes were visualized by immunoblotting withanti-V5 antibodies. Immunoreactive bands recognized by anti-V5 antibodies were observed in the anti-FGFR1 immunopre-cipitates from cells expressing mSefECTM but not mSe-fECpTM (Fig. 4). These data suggest that the transmembranedomain of Sef mediated, at least in part, the association betweenFGFR1 and mSef. In addition, since mSefECTM co-immuno-precipitated with both unstimulated and FGF-stimulatedFGFR1 with equal efficiency, these data indicated that theassociation between mSefECTM and FGFR1 was independentfrom ligand-induced FGFR1 activation.

3.5. The over-expression of Sef or Sef subdomains in HUVECinduces apoptosis

Since the survival of human umbilical vein endothelial cells(HUVEC) in vitro is dependent upon steady-state activation ofFGF signaling [35], we explored the effect of over-expressionof either mSefFL or its truncated mutants, SefICpTM andSefECTM, on HUVEC survival. We transduced HUVEC withthe indicated recombinant adenoviruses, and monitored cellviability over different periods of time. As judged by GFPfluorescence, the cells were transduced with an efficiencyexceeding 90% (data not shown). Significant cell death wasobserved in the population of Sef-expressing cells at 48h aftertransduction (Fig. 5A). Immunoblot analysis performed withantibody against cleaved caspase-3 indicated that cell death wasdue to apoptosis (Fig. 5C). The amount of viable cells at 72 hafter transduction was quantified by using an MTT cell viabilityassay. The results of three independent experiments showed thatat 72 h, the viability of HUVEC populations expressingmSefFL, mSefICpTM, and mSefECTM was reduced relativeto AdGFP-transduced cells by an average of 87%, 55%, and85%, respectively (Fig. 5B).

We previously reported that prolonged over-expression(∼36 h) of mSefFL or mSefICTM but not mSefECTM inHEK293T cells induces apoptotic cell death [36]. Our presentdata of comparable apoptotic effects induced by mSefFL ormSefECTM over-expression in HUVEC suggest that themSefECTM-mediated reduction of cell viability is most likelya result of the specific inhibition of FGF signaling in HUVEC,rather than FGF-independent apoptotic activity of mSefECTM.In order to gain more insight into the mechanisms underlyingSef-mediated reduction of HUVEC viability, we investigatedthe steady-state level of activation of several MAP kinases inAdSefFL, AdSefICpTM, and AdSefECTM-transduced cells.

Immunoblot analysis of cell lysates with antibodies, whichspecifically recognized activated forms of ERK1/2, JNK or p38MAPK demonstrated that at 48 h after transduction, theactivation of ERK1/2 or p38 MAPK was significantly reducedin mSefFL- and mSefECTM-expressing cells (Fig. 5C). Thelevel of ERK1/2 phosphorylation in HUVEC transduced withmSefICpTM adenovirus was reduced, although to a lesserextent, while the level of p38 MAPK activation remainedunchanged in these cells. The steady-state activation of JNKMAPK was not affected in any case (Fig. 5C).

Since both the presence of FGF in culture medium andsteady-state activation of MEK–ERK signaling pathway areimportant survival factors for HUVEC ([35], and data notshown), we next attempted to determine whether Sef-mediatedinhibition of ERK1/2 activation in HUVEC was due to aspecific inhibition of FGF signaling. First, we determinedwhether the presence of FGF in growth medium contributed to

1964 D. Kovalenko et al. / Cellular Signalling 18 (2006) 1958–1966

ERK1/2 activation in HUVEC. We cultured HUVEC in mediacontaining 10% serum with or without FGF2 for 24 h, and celllysates were subjected to immunoblot analysis with antibodiesagainst activated forms of ERK1/2 and MEK1/2 (Fig. 6A). Thisanalysis demonstrated that the presence of FGF in a culturemedium significantly contributed to a steady-state activation ofMEK–ERK signaling pathway in HUVEC. This conclusionwas supported by immunoblot analysis of levels of MEK1/2and ERK1/2 activation in HUVEC over-expressing a dominant-negative form of FGFR1 (dnFGFR) (Fig. 6B). Notably, over-expression of mSefECTM decreased MEK1/2 and ERK1/2phosphorylation to a similar extent as dnFGFR. This maysuggest that this effect can be attributed to specific inhibition ofFGF-dependent rather than serum-dependent ERK1/2 activa-tion in HUVEC. In order to determine whether mSefECTM actsas a specific inhibitor of FGF-induced ERK1/2 activation inHUVEC, we stimulated quiescent cells over-expressing mSe-fECTM with FGF2 (10 ng/ml) or serum (10%); subsequentimmunoblot analysis with anti-p-ERK1/2 antibody revealedthat over-expression of mSefECTM inhibited ERK1/2 activa-tion induced by stimulation with FGF2, while it had no effect onserum-induced ERK1/2 activation (Fig. 6C).

3.6. mSefECTM inhibits the function of a constitutivelyactivated form of FGFR1, FGFR1-C289R, but not FGFR1-K562E

To gain further insight into the mechanism by whichmSefECTM exerts its inhibitory function, we co-transfectedHEK293T cells with ERK2-GFP and either of two forms ofconstitutively activated FGFR1, FGFR1-C289R or FGFR1-K562E. As previously reported [29], the constitutive activationof FGFR1-C289R, in contrast to FGFR1-K562E, requiresreceptor dimerization; therefore, it can be inhibited by anintracellular domain truncated dominant-negative form ofFGFR1 (dnFGFR). We hypothesized that mSefECTM acts ina manner similar to dnFGFR, and will inhibit ERK activationinduced by FGFR1-C289R but not FGFR1-K562E. Indeed,immunoblot analysis of cell lysates performed with antibodies

Fig. 7. mSefECTM inhibits ERK activation induced by a constitutively activeform of FGFR1, FGFR1-C289R, but not FGFR1-K562E. HEK293T cells weretransiently transfected with constitutively activated FGFR1-C289R (0.5 μg) orFGFR1-K562E (0.5 μg) with ERK2-GFP (0.5 μg), and either with mSefFL,mSefICpTM or mSefECTM (5 μg of each), followed by serum starvation for12 h. Cell lysates were subjected to immunoblotting with anti-p-ERK1/2 (p-ERK2-GFP), anti-ERK1/2 (ERK2-GFP), anti-FGFR1 (clone 5G11), and anti-V5 (Sef) antibodies.

against activated forms of ERK1/2 demonstrated that themSefECTM was able to inhibit FGFR-C289R but not FGFR1-K562E-induced ERK activation (Fig. 7). Inhibitory effects ofboth mSefFL and mSefICpTM were not affected by the type ofconstitutively activated FGFR1 used (Fig. 7).

4. Discussion

Sef was originally identified as a gene whose expression ispositively regulated by FGF signaling during early zebrafishdevelopment. Subsequent functional analysis revealed thatzebrafish Sef (zSef) provides a physiologically importantfeedback inhibition of FGF signaling during zebrafish embryo-genesis [23,24].

Although Sef has been described as an inhibitor of FGFsignaling [23,24], the mechanism(s) by which this inhibitionoccurs remains an unresolved issue. It was originally reportedthat zSef co-immunoprecipitates with FGFR1 and FGFR2,indicating that Sef may act at the level of FGFR [23]; however,it was also reported that zSef functions at or downstream ofMEK [24]. Several recently published reports concerning themechanism of inhibitory action of Sef support both possibilities[25–28,37]. The over-expression of human Sef (hSef) inHEK293T cells was found to inhibit FGF-induced ERKphosphorylation without affecting upstream MEK activation[27]. It was also shown that the presence of an intracellulardomain is required for hSef inhibitory activity [27]. Inagreement with this, there is a report demonstrating that over-expression of a cytosolic form of hSef, which contains asubstituted N-terminal signal sequence, antagonizes ERK butnot MEK activation induced by FGF or PDGF in NIH3T3 cells[28]. Furthermore, it was reported that hSef acts as a molecularswitch for ERK signaling by specifically blocking ERK nucleartranslocation without inhibiting its activity in the cytoplasm[37]. In contrast, other reports argue that Sef inhibits FGFsignaling by acting upstream of Ras and, most likely, at the levelof FGFR. In particular, a recent study of hSef in PC-12 andHEK293T cell lines showed that hSef suppresses ERKactivation induced by FGF in PC-12 cells, while it has noeffect on the signaling mediated by constitutively active MEKor Ras in both PC-12 and HEK293T cells [26]. In agreementwith this, we have previously reported that mouse Sef (mSef),when over-expressed in HEK293T cells, inhibits ERK activa-tion induced by a constitutively activated form of FGFR1 butnot by constitutively active Ras [25]. Furthermore, wedemonstrated that mSef associates with FGFR1 in co-immunoprecipitation assay, and that the intracellular portionof mSef mediates, in part, this interaction. Moreover, wereported that mSef exerts its inhibitory function by inhibitingligand-induced phosphorylation of FGFR1 in NIH3T3 cells[25].

Although it is possible that Sef may act at different levels ofFGF signaling, our previously published results argue thatmSef-mediated inhibition of FGFR activation contributes, atleast in part, to overall inhibitory activity of mSef. Previously,we published data indicating that the mSef truncated mutantcontaining cytoplasmic and transmembrane domains of mSef

1965D. Kovalenko et al. / Cellular Signalling 18 (2006) 1958–1966

(mSefICTM) effectively binds to FGFR1; however, itsinhibitory effect on FGF-induced ERK activation was reducedwhen compared with full-length mSef [25]. These datasuggested a mechanism that may require additional domainsof Sef for full inhibitory activity. In the present study, weexamined the role of the mSef extracellular domain in mSef-mediated inhibition of FGF signaling. Over-expression studiesin NIH3T3, HEK293T, and HUVEC cultures indicated that theextracellular and transmembrane domains of mSef contributedto mSef inhibitory function. In fact, we demonstrate that theintracellular domain deletion mutant of mSef, mSefECTM,inhibited FGF signaling in NIH3T3 cells through inhibitingligand-induced FGFR1 phosphorylation. Because this inhibi-tion was not accompanied by a reduction in the total level ofFGFR1, it is unlikely that a decrease in FGFR1 stabilitycontributed to mSefECTM-mediated inhibitory function. Thesubstitution of the Sef transmembrane domain in mSefECTMprotein with a PDGFR transmembrane domain completelyabrogated its ability to interact with FGFR1 and to inhibit FGF-induced ERK activation. A similar substitution in themSefICTM truncation mutant had no inhibitory effect on eitherits interaction with FGFR1 (data not shown) or its functionalactivity (Fig. 2B). Interestingly, we frequently noted a decreasein FGFR1 levels in co-immunoprecipitation experiments in293T cells with mSefICTM, suggesting that this mutant mayaffect FGFR1 stability or intracellular trafficking. These datasuggest that although the transmembrane domain of mSe-fECTM plays a crucial role in establishing an associationbetween mSefECTM and FGFR1, it does not contribute tooverall inhibitory activity of mSefECTM which, therefore, canbe attributed to the extracellular portion of the protein. Arecently described alternative splicing mechanism, whichreplaces a functional N-terminal signal sequence of Sef andconverts the transmembrane form of Sef into a cytosolic form[28], can be considered a physiological regulator of Seffunctions that are attributed to the Sef extracellular andtransmembrane domains. Such a regulatory mechanism mayoccur in vivo in thyroid tissue, testes, and primary endothelialcells, where the alternatively spliced isoform of Sef has beendetected [28]. Interestingly, mRNAs of both transmembrane andcytosolic isoforms of Sef were found to be highly expressed inHUVEC [28], and Sef mRNA expression in HUVEC can beinduced by FGF stimulation [27]. These data suggest that Sefmay serve as a feedback regulator of FGF signaling in HUVEC,and that its expression can be regulated at both thetranscriptional and post-transcriptional levels. Our studysupports the hypothesis that Sef acts as a negative regulator ofFGF signaling in HUVEC, and shows that the extracellular andtransmembrane domains of Sef contribute to this regulation. Inaddition, our data demonstrate a functional activity ofmSefECTM that inhibited both FGF signaling and FGF-dependent survival of HUVEC. These data support the notionthat alternative splicing of Sef mRNA affects Sef subcellularlocalization, and can be viewed as an important physiologicalmechanism regulating Sef functions in endothelial cells. Thefunction of the cytosolic splice variant in HUVEC is currentlynot known.

We previously reported that Sef plays a role in the regulationof several FGF-independent processes, in particular, in theactivation of the JNK signaling pathway and induction ofapoptosis [36]. We demonstrated that long-term (∼36 h) over-expression of mSefFL or its intracellular domain-containingmutant, but not mSefECTM or mSefECpTM induces asignificant level of apoptosis in HEK293T cells, and that amSef-mediated activation of TAK1-MKK4-JNK signalingcascade contributes to this process [36]. In order to minimizethis Sef-mediated apoptotic effect, the time-frame for most ofthe experiments presented study did not exceed 18h. Within thistime-frame, the level of apoptosis in Sef-expressing cells wasindistinguishable from the background (data not shown). In thiscontext, a noticeable decrease in cell viability observed afterlonger-term over-expression of mSefECTM in HUVEC seemsto be uniquely correlative to the well-known FGF signalingsteady-state activation requirement for HUVEC survival [35].In contrast to HEK293T cells in which the over-expression ofmSefFL or mSefICpTM was shown to induce JNK activation[36], such activation was not observed in HUVEC, suggestingthe presence of cell-type specific mechanisms regulating Sef-mediated JNK activation. Those data also indicate that Sef-mediated induction of apoptosis in HUVEC was independentfrom the activation of JNK pathway, and can be attributed toSef-mediated specific inhibition of FGF signaling, in particular,FGF-dependent activation of MEK–ERK pathway. In thiscontext, cell viability could be partially rescued in HUVECexpressing mSefECTM by the addition of VEGF to the growthmedium (data not shown).

We found that mSefECTM interacted with and inhibitedligand-induced FGFR1 activation; however, we were unable todemonstrate that mSefECTM interfered with FGF2 binding toFGFR1 on the cell surface (data not shown). We hypothesizedthat mSefECTM may act in a similar fashion as dominant-negative FGFR (dnFGFR) by interfering with dimerization-dependent FGFR1 activation. Indeed, we have found that over-expression of mSefECTM in HEK293T cells was able to inhibitthe function of a constitutively activated form of FGFR1,FGFR1-C289R, while it had no effect on function of FGFR1-K562E whose activation occurs most likely due to conforma-tion change in the kinase domain, and does not require receptordimerization [29]. These results are consistent with previouslypublished data that the over-expression of dnFGFR is capable ofinhibiting FGFR1-C289R, but not FGFR1-K562E activity [29],and suggest that the mechanism(s) of Sef-mediated inhibition ofFGF signaling may include, in part, the ability of Sef to interferewith FGFR dimerization, although the interference with someconformational changes associated with activation of thisreceptor tyrosine kinase cannot be excluded.

5. Conclusions

In summary, our results demonstrate that the mSef extracel-lular and transmembrane domains contribute to mSef-mediatedinhibition of FGFR1 activation and, therefore, indicate thepotential structure–activity relationship between mSef subdo-mains in overall inhibitory function of mSef in FGF signaling.

1966 D. Kovalenko et al. / Cellular Signalling 18 (2006) 1958–1966

Acknowledgements

This work was supported by National Institutes of HealthGrants HL65301 (to R.F.) and COBRE grant RR15555 from theNational Center for Research Resources (to R.F.) We thank theDevelopmental Therapeutics Program of the NCI, NationalInstitutes of Health, for providing HUVECs. We thank LucyLiaw, Igor Prudovsky and other members of the Center forMolecular Medicine for critical review of this manuscript.

Appendix A. Supplementary data

Supplementary data associated with this article can be found,in the online version, at doi:10.1016/j.cellsig.2006.03.001.

References

[1] C.J. Powers, S.W. McLeskey, A. Wellstein, Endocr. Relat. Cancer 7 (2000)165.

[2] D.M. Ornitz, N. Itoh, Genome. Biol. 2 (2001) REVIEWS3005.[3] D.M. Ornitz, P.J. Marie, Genes Dev. 16 (2002) 1446.[4] K. Kannan, D. Givol, IUBMB Life 49 (2000) 197.[5] B.W. Draper, D.W. Stock, C.B. Kimmel, Development 130 (2003) 4639.[6] M. Sleeman, J. Fraser, M. McDonald, S. Yuan, D. White, P. Grandison, K.

Kumble, J.D. Watson, J.G. Murison, Gene 271 (2001) 171.[7] R. Friesel, T. Maciag, Thromb. Haemost. 82 (1999) 748.[8] P. Klint, L. Claesson-Welsh, Front. Biosci. 4 (1999) D165.[9] H. Larsson, P. Klint, E. Landgren, L. Claesson-Welsh, J. Biol. Chem. 274

(1999) 25726.[10] M.Mohammadi, C.A. Dionne, W. Li, N. Li, T. Spivak, A.M. Honegger, M.

Jaye, J. Schlessinger, Nature 358 (1992) 681.[11] K.G. Peters, J. Marie, E. Wilson, H.E. Ives, J. Escobedo, M. Del Rosario,

D. Mirda, L.T. Williams, Nature 358 (1992) 678.[12] H. Kouhara, Y.R. Hadari, T. Spivak-Kroizman, J. Schilling, D. Bar-Sagi, I.

Lax, J. Schlessinger, Cell 89 (1997) 693.[13] A. Aronheim, D. Engelberg, N. Li, N. Al-Alawi, J. Schlessinger, M. Karin,

Cell 78 (1994) 949.

[14] L. Bonfini, C.A. Karlovich, C. Dasgupta, U. Banerjee, Science 255 (1992)603.

[15] D. Bowtell, P. Fu, M. Simon, P. Senior, Proc. Natl. Acad .Sci. U. S. A. 89(1992) 6511.

[16] S.J. Leevers, H.F. Paterson, C.J. Marshall, Nature 369 (1994) 411.[17] S.G. Macdonald, C.M. Crews, L. Wu, J. Driller, R. Clark, R.L. Erikson, F.

McCormick, Mol. Cell. Biol. 13 (1993) 6615.[18] C.M. Crews, R.L. Erikson, Proc. Natl. Acad. Sci. U. S. A. 89 (1992) 8205.[19] E. Hulleman, J. Boonstra, Cell. Mol. Life Sci. 58 (2001) 80.[20] B.A. Ballif, J. Blenis, Cell Growth Differ. 12 (2001) 397.[21] M. Tsang, I.B. Dawid, Sci. STKE 2004 (2004) pe17.[22] R.T. Bottcher, N. Pollet, H. Delius, C. Niehrs, Nat. Cell. Biol. 6 (2004) 38.[23] M. Tsang, R. Friesel, T. Kudoh, I.B. Dawid, Nat. Cell Biol. 4 (2002) 165.[24] M. Furthauer, W. Lin, S.L. Ang, B. Thisse, C. Thisse, Nat. Cell Biol. 4

(2002) 170.[25] D. Kovalenko, X. Yang, R.J. Nadeau, L.K. Harkins, R. Friesel, J. Biol.

Chem. 278 (2003) 14087.[26] S. Xiong, Q. Zhao, Z. Rong, G. Huang, Y. Huang, P. Chen, S. Zhang, L.

Liu, Z. Chang, J. Biol. Chem. 278 (2003) 50273.[27] R.B. Yang, C.K. Ng, S.M.Wasserman, L.G. Komuves,M.E. Gerritsen, J.N.

Topper, J. Biol. Chem. 278 (2003) 33232.[28] E. Preger, I. Ziv, A. Shabtay, I. Sher, M. Tsang, I.B. Dawid, Y. Altuvia, D.

Ron, Proc. Natl. Acad. Sci. U. S. A. 101 (2004) 1229.[29] K.M. Neilson, R. Friesel, J. Biol. Chem. 271 (1996) 25049.[30] S. Hardy, M. Kitamura, T. Harris-Stansil, Y. Dai, M.L. Phipps, J. Virol. 71

(1997) 1842.[31] A. Fasbender, J. Zabner, M. Chillon, T.O. Moninger, A.P. Puga, B.L.

Davidson, M.J. Welsh, J. Biol. Chem. 272 (1997) 6479.[32] A.V. Miagkov, D.V. Kovalenko, C.E. Brown, J.R. Didsbury, J.P. Cogswell,

S.A. Stimpson, A.S. Baldwin, S.S. Makarov, Proc. Natl. Acad. Sci. U. S. A.95 (1998) 13859.

[33] R. Friesel, W.H. Burgess, T. Maciag, Mol. Cell. Biol. 9 (1989) 1857.[34] K. Mood, R. Friesel, I.O. Daar, J. Biol. Chem. 277 (2002) 33196.[35] S. Araki, Y. Shimada, K. Kaji, H. Hayashi, Biochem. Biophys. Res.

Commun. 168 (1990) 1194.[36] X. Yang, D. Kovalenko, R.J. Nadeau, L.K. Harkins, J. Mitchell, O.

Zubanova, P.Y. Chen, R. Friesel, J. Biol. Chem. 279 (2004) 38099.[37] S. Torii, M. Kusakabe, T. Yamamoto, M. Maekawa, E. Nishida, Dev. Cell 7

(2004) 33.