747Research Article

IntroductionExtracellular signaling is commonly mediated through mechanismsthat rely on protein-protein interactions and protein phosphorylation.The dynamics of protein-protein interactions are often dependenton the activation of a particular signal transduction pathway. Evena single protein can affect various cellular functions by interactingwith different partners in response to extracellular stimuli.

Recent advances in the development of molecular probes haveenabled the visualization of many cellular processes and thedetection of protein-protein interactions in living cells, advancestypified by the fluorescence resonance energy transfer (FRET)system (Förster, 1959), in addition to complementary methods usingfluorescence (Hu et al., 2002) or luminescence (Luker et al., 2004).These complementary methods designed to visualize protein-protein interactions have recently been employed in the screeningof interaction partners and the potential semi-high-throughputanalysis of small modulator molecules in protein complexes(Kerppola, 2006a; Kerppola, 2006b). In particular, luciferase-basedcomplementation imaging enables sensitive real-time monitoringand quantitative analysis of protein interactions under variouscellular conditions (Luker et al., 2004).

Integrins and their associated proteins mediate variousintracellular signaling pathways involved in cytoskeletal remodelingand cellular processes such as cell proliferation, survival anddifferentiation. Integrin-linked kinase (ILK), PINCH and parvin(s)form heterotrimeric complexes that function as important regulatorsof integrin-mediated signaling. The complex comprising ILK,PINCH and parvins has been implicated in mediating intracellular

signaling pathways through phosphorylation of the leadingdownstream molecule Akt (also known as protein kinase B; PKB)(Legate et al., 2006). ILK-dependent target phosphorylation islargely regulated by phosphoinositide 3-kinase (PI3K). Akt/PKBactivation requires phosphorylation of Thr308 by PI3K-dependentkinase-1 (PDK1) (Alessi et al., 1997; Williams et al., 2000) andSer473 by PDK2 [which is also known as hydrophobic motif kinase(HMK)] (Feng et al., 2004; Troussard et al., 2003). ILK is capableof controlling vascular endothelial growth factor (VEGF)transcription through Akt/PKB phosphorylation (Tan et al., 2004).ILK kinase activity is also stimulated by VEGF and other growthfactors, as well as by cell adhesion to the extracellular matrix(Attwell et al., 2003).

Parvins are a family of proteins involved in linking integrins andtheir associated proteins with intracellular pathways that regulateactin cytoskeletal dynamics and cell survival. It includesactopaxin/CH-ILKBP/a-parvin (ParvA), affixin/b-parvin (ParvB)and g-parvin (ParvG). ParvB localizes to focal adhesions and isinvolved in cell adhesion, spreading, motility and survival throughinteractions with its partners. ParvB accumulates and co-localizeswith ILK in heart and skeletal muscle (Bendig et al., 2006; Yamajiet al., 2001). It can inhibit ILK activity and reverse some oncogeniceffects in cancer cells (Mongroo et al., 2004). Furthermore, thephysiological interaction of ILK with ParvB is thought to beessential in maintaining cardiac contractility (Bendig et al., 2006).Nonetheless, the precise role of ParvB and the mechanismcontrolling the association of ILK and ParvB remain to bedelineated.

Functional molecular imaging of ILK-mediatedAkt/PKB signaling cascades and the associated roleof b-parvinMariko Kimura1,2,*, Takashi Murakami2,*, Shinae Kizaka-Kondoh3, Masayuki Itoh4, Keiji Yamamoto1, Yukihiro Hojo1, Makoto Takano4, Kazuomi Kario1, Kazuyuki Shimada1 and Eiji Kobayashi21Division of Cardiovascular Medicine, Department of Medicine, 2Division of Organ Replacement Research, Center for Molecular Medicine and4Division of Molecular Physiology and Biophysics, Department of Physiology, Jichi Medical University, Shimotsuke 329-0498, Tochigi, Japan3Department of Molecular Oncology, Kyoto University Graduate School of Medicine, Kyoto 606-8507, Japan*Authors for correspondence (mkimura@jichi.ac.jp; takmu@jichi.ac.jp)

Accepted 2 December 2009Journal of Cell Science 123, 747-755 © 2010. Published by The Company of Biologists Ltddoi:10.1242/jcs.052498

SummaryVisualization and quantification of the dynamics of protein-protein interactions in living cells can be used to explore the macromolecularevents involved in signal transduction processes. In this study, functional molecular imaging using a luciferase-based complementationmethod demonstrated how the integrin-linked kinase (ILK)-mediated protein complex controls downstream signals. The luciferasecomplementation assay showed that Akt1 preferentially binds to b-parvin rather than to ILK within the complex. Moreover, photonflux from the interaction between b-parvin and Akt1 increased following serum stimulation, and the b-parvin–Akt1 interaction wasdependent on phosphoinositide 3-kinase. Intriguingly, small interfering (si)RNA-mediated b-parvin knockdown increased photon fluxfrom the interaction between ILK and Akt1, leading to stabilization of hypoxia-inducible factor-1a and increased expression of vascularendothelial growth factor-A. These data from functional molecular imaging demonstrated that b-parvin plays a regulatory role in theILK-mediated Akt (also called protein kinase B) signaling cascades, suggesting that b-parvin might be a crucial modulator of cellsurvival.

Key words: b-parvin, Molecular imaging, Integrin-linked kinase, Hypoxia-inducible factor, Luciferase

Jour

nal o

f Cel

l Sci

ence

748

Herein, we demonstrate that luciferase complementation imagingprovides a useful strategy for the instantaneous monitoring ofprotein-protein interactions under serum stimulation. Employmentof this system enabled the detection of limited protein-proteininteractions, showing that ParvB preferentially interacts with Akt1,and inhibits the interaction between ILK and Akt1 under normalcell culture conditions. ParvB knockdown using small interfering(si)RNA increased VEGF expression, with concomitant stabilizedexpression of hypoxia-inducible factor 1-a (HIF-1a). Based on thesefindings, we propose that ParvB functions as an upstream modulatorof Akt/PKB through ILK, Akt/PKB and HIF-1a signaling.

ResultsFluorescence complementation assay for protein-proteininteractions involving ILK and associated proteinsIt has been demonstrated that ILK protein complexes play a crucialrole in regulating gene transcription and cell-cell adhesion. Molecularcomplementation methods were employed in an effort to detect real-time protein-protein interactions of ILK protein complexes withParvB, PINCH1 and Akt1 (Fig. 1A). A fluorescent (or luminescent)probe was split into N-terminal and C-terminal probe fragments, eachof which was attached to the N terminus or C terminus of the targetprotein using a linker. The optimal combination for each target protein-protein interaction leads to integration of the N-terminal and C-terminal probe fragments and light is emitted (Fig. 1A, right panel).To determine the optimal pairs of plasmids, four expression plasmidswere constructed for each target protein fused with fluorescent(monomeric Kusabira-Green: mKG) or luminescent (fireflyluciferase: Luc) N-terminal or C-terminal probe fragments using alinker (Fig. 1B). Protein interactions were investigated for structuralconformations comprising ILK and ParvB, ILK and PINCH1, ILKand Akt1, ParvB and Akt1, and PINCH1 and Akt1. As shown in Fig.1C, eight combinations per pair were tested to determine the optimalconformation in the living cell. The appropriate pairs of constructedplasmids were then co-transfected into 293T cells, which were theninspected using a fluorescent microscope under excitation light. Oneoptimal pair with respect to target protein interactions was selectedfrom among the eight combinations based on signal intensity. Theoptimal pairs for an mKG probe were N-ILK-linker-mKGC and N-ParvB-linker-mKGN (Fig. 1B: construct type 2 for ILK and constructtype 1 for ParvB), mKGC-linker-ILK-C and mKGN-linker-PINCH1-C (construct type 4 for ILK and construct type 3 for PINCH1), N-ILK-linker-mKGN and N-Akt1-linker-mKGC (construct type 1 forILK and construct type 2 for Akt1), N-ParvB-linker-mKGC andmKGN-linker-Akt1-C (construct type 2 for ParvB and construct type3 for Akt1). The numbers represent the same four types of plasmidsas in Fig. 1B. The plasmids encoded an unrelated protein and singleplasmids did not show substantial fluorescent signals (Fig. 1D, lowersix panels). Whereas the combination of ILK with ParvB or PINCH1gave steady fluorescent signals in the cytosol (Fig. 1D, upper leftpanels), the combination of ParvB with Akt1 yielded strongerfluorescent signals than those yielded by the combination of ILKwith Akt1 (Fig. 1D, upper right panels). The combination of PINCH1with Akt1 did not generate substantial fluorescent signals in this assay(data not shown). In an effort to confirm the interaction between ParvBand Akt1, we performed a GST pull-down assay. Although a directinteraction between ParvB and Akt had not been reported before thisstudy, the GST pull-down assay demonstrated that ParvB interactsdirectly with Akt1 (Fig. 1E). These results suggest that ParvBpreferentially interacts with Akt1 and that ParvB might serve as aconnecting molecule for ILK-mediated Akt/PKB signal transduction.

Luminescence complementation assay for protein-proteininteractions involving ILK and associated proteinsAlthough a variety of methodologies have been developed for theinvestigation of protein interactions (Kerppola, 2006a; Kerppola,2006b), the luciferase-based complementation assay systempotentially enables straightforward and real-time quantitativeanalysis in living cells (Luker et al., 2004; Paulmurugan et al., 2002).As with the fluorescence complementation assay, polypeptide ends,which can adopt a broad range of intermolecular orientations, mightbe problematic in terms of luciferase complementation imaging(Kerppola, 2006a; Kerppola, 2006b). Optimal luciferasecomplementation was investigated based on results derived fromfluorescence complementation assays (see Fig. 1). The optimal N-and C-terminal fragments of firefly luciferase cDNA (NLuc andCLuc, respectively) were isolated by PCR, in which NLuc and CLuccorrespond to firefly luciferase amino acids 2-416 and 398-550,respectively (Luker et al., 2004). Four fusion constructs weregenerated for each target protein using the NLuc and CLuc cDNAfragments (see Fig. 1B). Luciferase activity was then investigatedby examining target protein-protein interactions. Expression plasmidpairs were transiently transfected into 293T cells. For the controltransfection, only NLuc or CLuc was employed. To determine whichpair generated optimal luminescent signals, the relative luciferaseactivity for each target protein interaction was examined. As shownin Fig. 2A-D, the optimal pairs were as follows: NLuc-linker-ILK-C and CLuc-linker-ParvB-C (construct type 3 for ILK and constructtype 4 for ParvB), CLuc-linker-ILK-C and NLuc-linker-PINCH1-C (construct type 4 for ILK and construct type 3 for PINCH1),NLuc-linker-ILK-C and N-Akt1-linker-CLuc (construct type 3 forILK and construct type 2 for Akt1), and NLuc-linker-ParvB-C andCLuc-linker-Akt1-C (construct type 3 for ParvB and construct type4 for Akt1). Numbers represent the same four types of plasmids asin Fig. 1B. Single plasmids fused with only NLuc or CLuc cDNAdid not show substantial luminescent signals. In order to validateactual expression levels of ParvB and Akt1 protein, 2 mg of theconstructed plasmid DNA was transfected into 293T cells seededin 12-well plates. ParvB and Akt protein expression levels wereanalyzed by western blotting (supplementary material Fig. S1). N-terminally fused forms of Akt1 protein, NLuc-linker-Akt1-C andCLuc-linker-Akt1-C, were detected in two cleaved fragments. Ithas been demonstrated that Akt can be cleaved by recombinantcaspase 3 at aspartic acid residues 108 and 119, resulting in thegeneration of a 44 kDa pleckstrin-homology (PH) domain deficientfragment (Bachelder et al., 2001; Rokudai et al., 2000). Onefragment appeared to be the PH domain of Akt1 fused with a split-luciferase probe, which was detectable using anti-luciferaseantibody, and the other seemed to be the kinase domain of Akt1.These data indicated that the appropriate plasmid pair exclusivelyprovided optimal luminescent signals.

Real-time quantitative analysis of protein-proteininteractions using a luciferase-based luminescentcomplementation assayIn an effort to determine the kinetic pattern of the ILK-ParvB andILK-PINCH1 interactions, each optimal pair of plasmids wastransfected into 293T cells in a 24-well plate. Emitted photons weremeasured for 30 minutes at 30 second intervals in the presence ofD-luciferin using a NightOwl charge-coupled device camera.Maximal luminescence emitted from each well was observed at fiveto ten minutes. The photon count derived from the association ofILK with ParvB displayed a larger increase than that derived from

Journal of Cell Science 123 (5)

Jour

nal o

f Cel

l Sci

ence

749Role of b-parvin in ILK-Akt signaling

the association of ILK with PINCH1 at the maximumbioluminescent point (Fig. 3A,B).

Although ILK has been shown to bind to Akt (Persad et al., 2001),it had not been demonstrated that ParvB is associated with Akt1.

The kinetics pertaining to the interaction of Akt1 with ILK or ParvBwere therefore investigated. Optimal pairs of plasmids weretransfected into 293T cells under the same conditions. As shownin Fig. 3C,D, the increase in photon flux derived from the association

Fig. 1. Complementation imaging strategy to investigate protein interactions of ILK, ParvB,PINCH1 and Akt1. (A) Schematic representation of complementation methods used in this study. Afluorescent or luminescent probe was split into N- and C-terminal probe fragments that are fused with theN terminus or C terminus of either protein A or protein B (see Materials and Methods section). Whenprotein A and protein B interact appropriately, the two probe fragments are reconstituted and light signal isemitted. (B) Constructed plasmids are theoretically classified into four types according to the linkingposition of the two probes. Four expression plasmids for each target protein (ILK, ParvB, PINCH1 andAkt1) have the same basic backbone that contains the N- or C-terminal fragments of the fluorescent orluminescent protein probe. (C) Schematic representation of the appropriate complementation pairs ofplasmids. A potential eight combinations are employed to screen for optimum target protein-proteininteractions. Based on four types of plasmid structures, the optimum combination needs to be determinedas the indicated pair. (D) Optimized fluorescence signals were obtained for some complementation pairsof eight expression vectors pertaining to the target protein interactions. The optimized combination of ILKwith ParvB, ILK with PINCH1, ILK with Akt1 and ParvB with Akt1 provided fluorescent signals inliving 293T cells (upper panels). A related protein (NFkB1 p50 or RelA p65) with either ILK or ParvB,and the single plasmid alone (ILK, ParvB, PINCH1 and Akt1) did not give any fluorescent signals (lowerpanels). Scale bars: 100 mm. (E) GST pull-down assay. Glutathione-agarose beads coated with either GSTor GST-ParvB were mixed with His-tagged Akt1, washed and detected by CBB staining and westernblotting using anti-Akt antibody.

Jour

nal o

f Cel

l Sci

ence

750

of Akt1 with ParvB was much greater than that derived from theassociation of Akt1 with ILK, suggesting the preferential interactionof Akt1 with ParvB rather than ILK under living cell conditions.Maximum luminescent levels derived from the association of Akt1with ParvB were in the range 800-1600 counts/pixels (50- to 100-fold higher than that observed for Akt1 and ILK). Kinetic analysis

of Akt1-ILK complex formation showed an upward-bulging curve,with maximum luminescence levels being observed for 40-50minutes following the addition of D-luciferin. Control expressionusing intact luciferase reached a peak immediately following theaddition of D-luciferin (supplementary material Fig. S2). Theseresults indicate that real-time complementary imaging can reveal

Journal of Cell Science 123 (5)

Fig. 2. Relative luciferase activity in eight luciferasecomplementation pairs. The optimized pair (black bar) foundwas for the following combinations: (A) ILK and ParvB, (B)ILK and PINCH1, (C) ILK and Akt1, (D) ParvB and Akt1.Numbers in parentheses represent the same four types ofplasmids as in Fig. 1B.

Fig. 3. Kinetic luminescence signals of the optimal luciferasecomplementation pair in 293T cells. (A) ILK and ParvB, (B)ILK and PINCH1, (C) ILK and Akt1, (D) ParvB and Akt1. Thecomplementation imaging assay was performed under the sameconditions pertaining to panels A and B, or C and D. Withrespect to Akt1-ILK and Akt1-ParvB complex formation,representative pseudo-color images (four-part) of a 24-well plateshowed maximal luminescence signals 40 minutes following theaddition of D-luciferin (upper panels). Notably, the combinationof ParvB with Akt1 provided stronger signals than thecombination of ILK with Akt1.

Jour

nal o

f Cel

l Sci

ence

751Role of b-parvin in ILK-Akt signaling

characteristics unique to each target protein-protein interaction, inthis case suggesting that ParvB might be an important connectingmolecule in the ILK-Akt/PKB signaling pathway.

Functional complementation imaging demonstratedParvB-Akt1 interaction following serum stimulationIn an effort to glean insight into the role of ParvB, we set out todetermine whether the ParvB-Akt1 interaction was modifiedfollowing serum-triggered induction. As depicted in Fig. 4A,representative pseudo-color images based on the split luciferasesystem revealed that photon flux derived from the interaction ofParvB with Akt1 increased under serum-rich conditions anddiminished under serum-starved conditions or in the presence of aPI3K inhibitor. Consistent with these findings were the resultsobtained when investigating the kinetics of the ParvB-Akt1interaction under serum-starved conditions or in the presence of aPI3K inhibitor with serum at 1-minute intervals following theaddition of D-luciferin (Fig. 4B-D). The kinetic curves werelogarithmic in shape. Maximum photon flux for the ParvB-Akt1interaction for each condition was observed 20-40 minutes followingserum stimulation. Serum-rich treatment increased photon fluxderived from the interaction of ParvB with Akt1 (Fig. 4B). PI3Kinhibitors such as LY294002 and wortmannin suppressed serum-triggered induction (Fig. 4C,D). Employment of a dual-luciferaseassay also demonstrated that luciferase activity arising from theParvB-Akt1 interaction in lysate cells was similar to photon fluxin live cells (Fig. 4E). However, with respect to the ILK-Akt1interaction and a control plasmid encoding full-length fireflyluciferase, there was no significant change in luciferase activityunder serum-starved or PI3K inhibitor conditions with serum(supplementary material Fig. S3A,B). These results demonstrate thatserum stimulation increases photon flux derived from the interactionof ParvB with Akt1, suggesting that PI3K inhibitors attenuate theserum-stimulated interaction between ParvB and Akt1.

Interaction between ILK and Akt1 is affected by ParvBmRNA levelsBecause ILK-dependent phosphorylation is regulated by PI3K, itis possible that the inhibition of PI3K activity decreases ILK activity,which subsequently impairs the phosphorylation of putative ILKsubstrates (Delcommenne et al., 1998). One reliable marker thatrelates to ILK activity concerns the phosphorylation levels ofAkt/PKB (Legate et al., 2006). Given the preferential associationof ParvB with Akt1, we set out to determine whether ParvB levelsregulate the interaction between ILK and Akt1. NIH3T3 cells weretransfected with a reporter plasmid encoding the optimal split-luciferase complementation pair ILK-ParvB (Fig. 5A) and thensubjected to siRNA-mediated ParvB knockdown. The knockdownefficiency of two types of mouse ParvB siRNAs (ParvB1 andParvB2) is shown in Fig. 5B. These siRNAs were transfected intoNIH3T3 cells stably overexpressing the complementation plasmidpair ILK-ParvB (see Fig. 5A) and the luminescence photon kineticswere evaluated (Fig. 5B). ParvB knockdown using siRNA decreasedphoton emission associated with ILK-ParvB association (siRNAParvB2 was able to reduce approximately 50% of the photonsderived from the interaction of ParvB with ILK). We then evaluatedthe extent of ILK-Akt1 complex formation with loss of ParvB usingboth luciferase complementation imaging and a co-immunoprecipitation assay. 293T cells were transfected with theoptimal complementation plasmid pair ILK-Akt1 and then subjectedto ParvB knockdown using siRNA ParvB2. As depicted in Fig. 5C,

we investigated the kinetics of photons derived from the interactionof ParvB with Akt1 under ParvB knockdown conditions at 1-minuteintervals following serum stimulation. Interestingly, siRNA ParvBled to an increase in photon flux associated with the ILK-Akt1interaction. To provide further evidence showing that ParvB mightregulate the ILK-Akt interaction, we performed a co-immunoprecipitation assay. 293T cells were transfected with theluciferase complementation plasmid pair ILK-ParvB (Fig. 5A) andthen subjected to ParvB knockdown using siRNA ParvB2 (Fig. 5B).Overexpressed ILK and ParvB were precipitated with anti-phospho-Akt (Ser473) antibodies and precipitates were analyzed by westernblotting using antibodies against ILK, ParvB, total-Akt and phospho-Akt (Ser473). As shown in Fig. 5D, decreased ParvB protein levelsresulted in increased interaction between ILK and phospho-Akt(Ser473). These results suggest that ParvB might act as a potentialmodulator of the signal transduction associated with the ILK-Akt/PKB complex.

Fig. 4. Kinetics of luciferase complementation imaging for ParvB-Akt1complex formation under different conditions. (A) Representative pseudo-color luminescence images emitted by the interaction between ParvB and Akt1under either serum-starved conditions or in the presence of PI3K inhibitor(LY294002 or wortmannin) 20 minutes following serum stimulation. (B) Kinetic interaction between ParvB and Akt1 under serum-starvedconditions following serum stimulation. 1% FCS versus 1% FCS plus serum. (C,D) Kinetic interaction between ParvB and Akt1 in the presence of PI3Kinhibitors following serum stimulation: (C) 20 mM LY294002; (D) 10 nMwortmannin. 1% FCS plus serum versus 1% FCS plus serum plus PI3Kinhibitor. (E) Relative luciferase activity for the interaction between ParvB andAkt1 20 minutes post-serum stimulation with or without PI3K inhibitors(LY294002 or wortmannin) All data are presented as a mean ± s.d. determinedfrom the analysis of more than three independent experiments. NS, notstatistically significant, **P<0.01, *P<0.05, ANOVA followed by post hoctests or Student’s t-test.

Jour

nal o

f Cel

l Sci

ence

752

ParvB knockdown increased HIF-1 and VEGF-Aexpression levelsWe showed above that ParvB protein levels regulate the ILK-Akt1interaction following serum stimulation. It has been demonstratedthat HIF-1a is an important downstream effector that acts throughthe PI3K-Akt/PKB signaling pathway (Jiang et al., 2001).Furthermore, HIF-1a can regulate the expression of VEGF(Forsythe et al., 1996; Liu et al., 1995). In an effort to examine theeffect of ParvB downregulation on downstream targets of the ILK-Akt signaling pathway, HIF-1a and VEGF protein expression levelswere investigated following siRNA-mediated ParvB knockdown.ParvB is ubiquitously expressed, but enriched in heart and skeletalmuscle (Bendig et al., 2006; Sepulveda and Wu, 2006). Wetransfected rat ParvB-specific siRNA into rat cardiomyocytes andconfirmed the knockdown effect of ParvB (Fig. 6A). Western blotanalysis demonstrated increased expression levels of endogenousHIF-1a and VEGF-A with ParvB knockdown in rat cardiomyocytes.HIF-1a and VEGF-A expression levels were quantified andstatistically analyzed (Fig. 6B,C).

In an effort to further investigate the increase in HIF-1a expressioninduced by siRNA ParvB, another reporter plasmid was employedcontaining oxygen-dependent degradation (ODD)-luciferase, whichcomprises the ODD domain of HIF-1a fused to luciferase. Becausethe stability of HIF-1a is tightly regulated through the ODD domain(Harada et al., 2002), luciferase activity associated with the fusionprotein reflects stabilized HIF-1a levels. ODD-luciferase cDNA wassubcloned into the pcDNA3.1 expression plasmid (Fig. 6D) andtransfected into NIH3T3 cells. ODD-luciferase-expressing stabletransformants were obtained following G418 selection. We thenintroduced siRNA ParvB1 or ParvB2 into the ODD-luciferase-

expressing NIH3T3 transformants. As shown in Fig. 6E, ParvBknockdown resulted in a marked increase in the expression of ODD-luciferase. Moreover, the hypoxia mimetics CoCl2 anddesferrioxamine (DFO) also increased ODD-luciferase-derivedphoton flux in the NIH3T3 transformants (Fig. 6F). To further confirmthe requirement for ILK in the induction of HIF-1a stability underloss of ParvB, both siRNA ParvB1 and siRNA ILK were introducedinto ODD-luciferase-expressing NIH3T3 transformants. As shownin Fig. 6G, knockdown using both ParvB and ILK did not increaseODD-luciferase-derived photon flux in the NIH3T3 transformants(Fig. 6G).

Taken together, these results indicate that decreased levels ofParvB stabilize HIF-1a in the presence of ILK, suggesting thatParvB downregulation might mimic hypoxic conditions through theILK-Akt/PKB signaling pathway.

DiscussionComplementation strategies using an imaging probe withappropriate protein reconstitution enable visualization of steady-state complexes formed between protein pairs. One of the benefitsassociated with the use of these techniques is the exclusion of certainsecondary effects or potential artifacts caused by cell lysis(Kerppola, 2006a; Kerppola, 2006b). Of the availablecomplementation strategies, luciferase-based luminescencecomplementation imaging can be employed as a facile and broadlyapplicable approach (Luker et al., 2004). In this study, wedemonstrated two complementary methods employing fluorescentor luminescent protein probes. Both imaging methods required thateight combinations of reporter plasmids be examined to determineoptimal signal gain from the target protein-protein interaction.

Journal of Cell Science 123 (5)

Fig. 5. Interaction between ILK and Akt1 underParvB knockdown conditions. (A) Schematicrepresentation of one expression vector encoding theoptimal luciferase complementation pair ILK andParvB. (B) ParvB knockdown by siRNA-inhibitedluminescence signals in NIH3T3 cells overexpressingthe optimal luciferase complementation pair ILK andParvB. Representative pseudo-color images (triplicate)of a 24-well plate (the upper panel) and itsquantification (the lower panel). (C) 293T cells weretransfected with the optimal luciferase complementationplasmid pair ILK and Akt1, and then treated with ParvBsiRNA. The graph represents photon-derived kinetics ofthe interaction between ILK and Akt1 following serumstimulation (left panel). Representative pseudo-colorluminescence images (right panel) of a 12-well plate 40minutes post-stimulation (broken line). (D) The plasmidencoding the optimal luciferase complementation pairILK and ParvB (A) was transfected into 293T cells, andcells were then treated with ParvB siRNA. Cell lysateswere precipitated (IP) using anti-phospho-Akt (Ser473)antibodies and control IgG antibody, and then subjectedto western blotting (WB) using anti-ILK, anti-ParvB,anti-Akt and anti-phospho-Akt (Ser473) antibodies. Alldata are presented as a mean ± s.d. (**P<0.01, ANOVAfollowed by post hoc tests or Student’s t-test).

Jour

nal o

f Cel

l Sci

ence

753Role of b-parvin in ILK-Akt signaling

Although fluorescence complementation enabled visual observationof the actual protein interactions, it might not necessarily allowcomparison of protein-protein interactions in living cells under themicroscope. By contrast, luminescence complementation enablescomparison of even weak protein interactions under almost real-time conditions by quantifying luminescent signals with a highsignal-to-noise ratio. Furthermore, employment of the luciferasecomplementation strategy allows the observation of rapid changesin target protein interactions in subcellular compartments undervarious conditions. In fact, our extensive complementary studiesdemonstrated that ParvB preferentially binds to Akt1 rather thanILK under living cell conditions. Moreover, luciferasecomplementation imaging revealed that real-time changes in photon-based kinetics associated with the ParvB-Akt1 interaction wereconsistent with results obtained following the use of extracellularstimuli, such as growth factors and associated signal inhibitors.Thus, employing the complementation strategy with appropriateprobes could provide unique findings that reflect real-time cellularresponses to external stimuli. Based on the current studies,employment of functional molecular imaging using complementarymethods promises to be a beneficial strategy for the exploration ofmolecular mechanisms pertaining to signal transduction.

The ILK-associated protein complex is profoundly involved inaltering the flux of the PI3K-Akt/PKB signaling pathway (see Fig.7). ILK-dependent phosphorylation is regulated in a PI3K-dependentmanner. PI3K inhibitors reduce ILK activity and impair thephosphorylation of putative ILK substrates in cell culture

(Delcommenne et al., 1998). ParvB is a binding partner of ILK andthe CH2 domain of ParvB is phosphorylated by ILK. These findingssuggest that ParvB probably plays a role in actin cytoskeletonremodeling and cell spreading with its binding partners, a-actininand aPIX (Mongroo et al., 2004; Yamaji et al., 2004; Yamaji et al.,2001). One new finding in our study showed that the real-time photon-based kinetics of the ParvB-Akt interaction correlated with resultsobtained following the use of extracellular stimuli, such as thepresence of serum with or without PI3K inhibitors (Fig. 4). Moreover,decreased levels of ParvB protein were associated with a markedincrease in photons derived from the ILK-Akt interaction (Fig. 5C,D).Although the interaction between ParvB and Akt had not beenreported before this study, ParvB might cooperate with Akt and playa regulatory role in ILK-Akt complex formation (Fig. 7).

The role of ParvB in relation to Akt/PKB downstream effectorshad not previously been sufficiently addressed. In this study, wedemonstrated that siRNA-mediated ParvB knockdown increased HIF-1a and VEGF expression in rat cardiomyocytes (Fig. 6B,C). Inparticular, we showed that the phenomenon was correlated with HIF-1a stabilization. In our luciferase-based real-time imaging, ODD-luciferase expression levels increased with siRNA-mediated ParvBknockdown, firmly supporting the notion of HIF-1a stabilization (Fig.6E). Furthermore, the hypoxia mimetics CoCl2 (100 mM) and DFO(20 mM) increased luminescent signals in ODD-luciferase-overexpressing NIH3T3 cells. The effect of ParvB knockdown wasthe same as that of hypoxia mimetic agents. However, knockdownusing both ParvB and ILK siRNAs did not increase luminescent

Fig. 6. Stabilization of HIF-1 and increasedexpression of VEGF-A under loss of ParvB.(A) Quantitative real-time RT-PCR for ParvB in ratcardiomyocytes following treatment with rat ParvB-specific siRNA. (B) siRNA-mediated ParvB knockdownincreased protein expression levels of HIF-1a and VEGF-A in cardiomyocytes following serum stimulation.(C) Quantitative analysis (using ImageJ software) ofendogenous HIF-1a (left panel) and VEGF-A (right panel)expression levels from the siRNA-mediated ParvBknockdown experiments shown in B. (D) Schematicrepresentation of the reporter plasmid construct encoding afusion protein, ODD-luciferase, which comprises part ofthe ODD domain of HIF-1a fused to luciferase (Harada etal., 2002). (E) siRNA-mediated ParvB knockdownincreased photon count in NIH3T3 cells overexpressingODD-luciferase. (F) Hypoxia mimetics CoCl2 (100 mM)and DFO (20 mM) increased luminescent signals inNIH3T3 cells overexpressing ODD-luciferase.(G) Double-knockdown using ParvB- and ILK-specificsiRNAs did not increase ODD-luciferase-derived photonsin the NIH3T3 transformants. All data are presented as amean ± s.d. determined from the analysis of more thanthree independent experiments (**P<0.01, *P<0.05,ANOVA followed by post hoc tests or Student’s t-test).

Jour

nal o

f Cel

l Sci

ence

754

signals in ODD-luciferase-overexpressing NIH3T3 cells (Fig. 6G).These results suggest that ParvB expression levels potentially controlHIF-1a levels in the presence of ILK and that HIF-1a levelssubsequently alter the expression of target genes such as VEGF-A.Taken together with the results presented here, one possible scenariocan be outlined (Fig. 7): growth factors trigger ILK-Akt/PKBsignaling activation and consequently ParvB could be phosphorylatedby activated ILK, leading to increased interaction between ParvBand Akt. In the normal and physiological state, ParvB might act asa modulator to prevent the transmission of excess signals, such asthose derived from growth factors, in the ILK-Akt/PKB signalingcascade (Fig. 7A). However, without regulation of ParvB, excesssignals might be transmitted in an uncontrolled manner to downstreamtargets such as HIF1-a or VEGF (Fig. 7B). The notion of ParvBacting as a modulator might be supported by the finding that ParvBis downregulated in breast cancer (Mongroo et al., 2004).

With respect to the role of ParvB in ILK–Akt/PKB–HIF-1asignaling, it has been demonstrated that loss of ParvB expressionupregulates ILK activity in tumors (Mongroo et al., 2004) and thatILK activation triggers a cascade of Akt/PKB phosphorylation eventsin tumor cells (Majumder et al., 2004). Because it has been reportedthat transcriptional responses to Akt/PKB activation are involved inthe mTOR-dependent regulation of HIF-1a in an Akt-driven prostateintraepithelial tumor, ParvB might be involved in HIF-1a stabilizationvia ILK-Akt/PKB signaling in tumor cells. Thus, it is believed thatour results are consistent with the role of ParvB in tumor cells.

Although further investigations might be necessary given thesefindings, employment of the complementary protein imagingstrategy enabled the visualization and quantification of protein-protein interactions. This approach allowed the detection of real-time alterations in ILK-Akt signaling under conditions involvingextracellular stimulation of living cells. These compelling datademonstrate that ParvB plays a regulatory role in ILK-Akt/PKB-HIFa signaling.

Materials and MethodscDNAs and plasmid constructionFor the isolation of ILK, ParvB, PINCH1 and Akt1 cDNA (GenBank accession numbers:NP_034692, NM_133167, NM_026148 and NM_009652, respectively), total RNA wasextracted from the heart of BALB/c mice using ISOGEN (Nippon Gene, Tokyo, Japan)following the manufacturer’s instructions. 1 mg of total RNA was reverse transcribedusing random hexanucleotide primers and SuperScript III (Invitrogen Carlsbad, CA)according to the manufacturer’s instructions. The entire coding sequence of these cDNAswas amplified by PCR using appropriate cloning primers (supplementary material TableS1). PCR was performed using EX Taq polymerase (Takara Bio, Otsu, Japan) as follows:94°C for 2 minutes, 59°C for 1 minute, 72°C for 2 minutes, followed by 30 cycles of

94°C for 30 seconds, 62°C for 30 seconds and 72°C for 1 minute, with a final extensionstep at 72°C for 5 minutes. PCR products were cloned into 2.1 TOPO vector (Invitrogen,Carlsbad, CA). To detect KG fluorescent signals formed by the protein-proteininteractions, ILK, ParvB, PINCH1 and Akt1 cDNAs were subcloned into phmKGN-MC, phmKGC-MC, phmKGN-MN and phmKGC-MN plasmids of the Fluo-chase kit(Medical & Biological Laboratories, Woburn, MA), which included the Kozakconsensus sequence.

For the split firefly luciferase complementation assay, optimal pairs of N- and C-terminal cDNA fragments were amplified by PCR from pGL4.10 plasmid DNA(Promega, Madison, WI). The N- or C-terminal cDNA fragment was inserted intothe pcDNA3.1 expression vector (Invitrogen). ILK, ParvB, PINCH1 and Akt1 cDNAswere fused in-frame to the N-terminal luciferase fragment (NLuc) or the C-terminalluciferase fragment (CLuc) containing a linker sequence.

Both ParvB cDNA fragments comprising N-terminal CLuc (CLuc-linker-ParvB)and ILK with N-terminal NLuc (NLuc-linker-ILK) were inserted into the pILES2-EGFP expression vector (Clontech Laboratories, Palo Alto, CA). An internalribosomal entry site (IRES) sequence was inserted upstream of the start site of theNLuc-linker-ILK cDNA.

All PCR primer sequences for cDNA cloning and subcloning are shown insupplementary material Table S1.

Cell culture and transfectionNIH3T3 and 293T cells were maintained in Dulbecco’s Modified Eagle’s Medium(DMEM; Sigma Chemical Co., St Louis, MO) supplemented with 10% heat-inactivated fetal calf serum (FCS), 1 mM sodium pyruvate, 10 mM HEPES buffer,2 mM glutamine, 100 units/ml penicillin G and 100 mg/ml streptomycin sulfate. Heartsfrom Wistar rat neonates were removed under a microscope, minced and digestedfour times at 37°C using a mixture of 1 mg/ml collagenase (Worthington BiochemicalCo, Lakewood, NJ) and Hank’s solution. Cardiac cells were neutralized and washedby low-speed centrifugation in 10% FCS-DMEM. After enzymatic digestion of theirtissues, a cardiomyocyte fraction was prepared by Percoll gradient centrifugation(GE Healthcare). Cardiomyocytes were plated on wells pre-incubated with FCS.

Transient transfection was performed with an 80% confluent cell culture of 293Tand NIH3T3 cells using Lipofectamine 2000 or Lipofectamine LTX (Invitrogen)according to the manufacturer’s instructions. For serum stimulation, transfected cellswere serum starved in 1% FCS for 18-20 hours and followed by FCS additionimmediately before the assay. For PI3K inhibition, cells were pretreated withLY294002 (Sigma) or wortmannin (Sigma) at the concentration specified in the figurelegends for 2 hours before the assay.

For the isolation of stable transformants that express the optimal complementationpair ILK and ParvB with EGFP (see Fig. 5A) and ODD-luciferase (Harada et al.,2002) (see Fig. 6D), NIH3T3 cells were grown in 10% FCS-DMEM containing 1000mg/ml G418 at 2 days post-transfection. Following 21 days of G418 selection,neomycin-resistant colonies were isolated.

For the siRNA experiments, ParvB-specific siRNA duplexes, ILK-specific siRNAduplexes and an unrelated siRNA duplex as a control were purchased from Invitrogen.The siRNA transfection was performed using Lipofectamine 2000 (Invitrogen)according to the manufacturer’s instructions. The final siRNA concentration was 20-40 nM.

Complementation imaging assayThe fluorescence complementation assay was performed using the Fluo-chase Kit(Medical & Biological Laboratories). 293T cells were co-transfected using the split-mKG vector system according to the manufacturer’s instructions. Cells were inspectedand photographed at 36-48 hours post-transfection using a fluorescent microscopeequipped with a UV light source (excitation 470-495 nm; emission 510-550 nm).

Journal of Cell Science 123 (5)

Fig. 7. Schematic representation of the role ofParvB in the ILK-Akt/PKB signaling pathway. (A) The normal and physiological state. ParvB mightact as a modulator to prevent the transmission ofexcess signals. (B) The ParvB-depleted state. Withoutregulation of ParvB, excess signals might betransmitted in an uncontrolled manner to downstreamtargets.

Jour

nal o

f Cel

l Sci

ence

755Role of b-parvin in ILK-Akt signaling

For the luciferase-based complementation imaging, firefly luciferase cDNA wassplit into two fragments. 293T cells were co-transfected with the appropriate eightpairs of constructed plasmids (see Fig. 1C). Cells were plated and then transientlyco-transfected with the appropriate pairs of plasmids using Lipofectamine 2000(Invitrogen). At 36-40 hours after transfection, photon flux from the cells wasmeasured using the NightOWL imager at 1-minute intervals following the additionof D-luciferin (potassium salt; Biosynth, Postfach, Switzerland). Cells were alsoassayed at 36-40 hours following transfection using the dual-luciferase reporter assaysystem (Promega) according to the manufacturer’s instructions.

Immunoprecipitation and western blotting293T cells were transfected with the constructed plasmid encoding the optimalluciferase complementation plasmid ILK-ParvB. Cells were lysed with a lysis buffer(1% Triton X in 50 mM HEPES, pH 7.4, containing 150 mM NaCl, 1 mM Na3VO4,50 mM NaF and protease inhibitors). Cell lysates were incubated overnight at 4°Cwith anti-phospho-Akt (Ser473) antibodies (Cell Signaling Technology, Beverly, MA)and then mixed with protein-G agarose beads (GE Healthcare, Piscataway, NJ).Precipitates were washed four times and then subjected to western blotting.

For western blot analysis, cells were lysed 2-3 days following transfection with alysis buffer (50 mM Tris-HCl, pH 7.4, 1% NP-40, 0.25% sodium deoxycholate, 0.1%SDS, 150 mM NaCl, 1 mM EDTA, 1 mM Na3VO4, 1 mM NaF and proteaseinhibitors). Cell lysates were subjected to 5-20% SDS-PAGE and then transferredonto polyvinylidene difluoride membranes. Blots were blocked using 2% bovine serumalbumin (Sigma) or 5% ECL blocking agent (GE Healthcare) in TBST buffer (20mM Tris-HCl, 150 mM NaCl and 0.1% Tween 20). Membranes were then incubatedovernight at 4°C with target primary antibodies in the blocking buffer at a dilutionrecommended by the manufacturer [phospho-Akt (Ser473), Akt and ILK: CellSignaling Technology, Beverly, MA; HIF-1a: Chemicon, Temecula, CA; VEGF-A:Immuno-Biological Laboratories Co, Gunma, Japan; luciferase: Promega; ParvB andGAPDH: Santa Cruz Biotechnology Inc, Santa Cruz, CA]. Membranes were washedextensively in TBST buffer and then incubated with Clean-Blot IP Detection ReagentHRP (Thermo Fisher Scientific Inc., Rockford, IL), anti-rabbit IgG (Cell SignalingTechnology), anti-goat IgG (Santa Cruz Biotechnology) or protein A (Bio-Rad,Hercules, CA) for 1 hour in TBST buffer. Membranes were finally washed with TBSTbuffer and signals were visualized using ECL plus (GE Healthcare).

Expression and purification of recombinant proteinsThe ParvB fragment generated by PCR was cloned between the BamHI and EcoRIrestriction sites of pGEX-6P-1 (GE Healthcare). GST-ParvB and GST were expressedin Escherichia coli BL21 (DE3) and purified using GST Bind Kits (Novagen,Darmstadt, Germany) according to the manufacturer’s instructions.

GST pull-down assayFor the pull-down assay, 5 mg of either GST or GST-ParvB mixed with 1 mg His-tagged Akt1 (Invitrogen) were bound to glutathione-coated agarose beads in 500 mlof binding buffer (150 mM NaCl, 50 mM HEPES, 0.1% NP-40, pH 7.4, 1 mMdithiotreitol) for 2 hours at 4°C. Bound proteins were washed three times and elutedwith SDS sample buffer, separated by SDS-PAGE, and detected by western blottingand Coomassie brilliant blue (CBB) staining.

Quantitative real-time RT-PCRTotal RNA was isolated using ISOGEN (Nippon Gene Co.) according to themanufacturer’s instructions. 1 mg of total RNA was reverse transcribed with oligo-dT primers using SuperScript III (Invitrogen) according to the manufacturer’sinstructions. ParvB gene expression was quantified using iQ SYBR Green Supermix(Bio-Rad). Each assay was performed in triplicate. ParvB transcript levels weredetermined as expression levels relative to 18S rRNA. The following primers wereused: rat ParvB, 5�-TTGTCCCCCTTCACAACTTC-3� and 5�-GTGTAAAG -GACCCGCAGAGT-3�; 18S rRNA, 5�-GTAACCCGTTGAACCCCATT-3� and 5�-CCATCCAATCGGTAGTAGCG-3�. Each experiment was performed three times withsimilar results.

StatisticsComparisons between multiple groups were performed using ANOVA followed bypost hoc tests or Student’s t-test. A P value less than 0.05 was considered significant.

We thank J. Hotta, Y. Ohide and M. Kumagai for excellent technicalassistance. This study was supported in part by a grant to T.M. fromthe Health and Labour Science Research Grants (Research on BiologicalResources), a Grant-in-Aid for Scientific Research from the JapanSociety for the Promotion of Science (JSPS; project number 20591326;2008), and a grant from the Strategic Research Platform for PrivateUniversities: matching fund subsidy from MEXT (2008-), and aResearch Award to JMU Graduate Student (M.K.).

Supplementary material available online athttp://jcs.biologists.org/cgi/content/full/123/5/747/DC1

ReferencesAlessi, D. R., James, S. R., Downes, C. P., Holmes, A. B., Gaffney, P. R., Reese, C. B.

and Cohen, P. (1997). Characterization of a 3-phosphoinositide-dependent protein kinasewhich phosphorylates and activates protein kinase Balpha. Curr. Biol. 7, 261-269.

Attwell, S., Mills, J., Troussard, A., Wu, C. and Dedhar, S. (2003). Integration of cellattachment, cytoskeletal localization, and signaling by integrin-linked kinase (ILK), CH-ILKBP, and the tumor suppressor PTEN. Mol. Biol. Cell 14, 4813-4825.

Bachelder, R. E., Wendt, M. A., Fujita, N., Tsuruo, T. and Mercurio, A. M. (2001).The cleavage of Akt/protein kinase B by death receptor signaling is an important eventin detachment-induced apoptosis. J. Biol. Chem. 276, 34702-34707.

Bendig, G., Grimmler, M., Huttner, I. G., Wessels, G., Dahme, T., Just, S., Trano, N.,Katus, H. A., Fishman, M. C. and Rottbauer, W. (2006). Integrin-linked kinase, anovel component of the cardiac mechanical stretch sensor, controls contractility in thezebrafish heart. Genes Dev. 20, 2361-2372.

Delcommenne, M., Tan, C., Gray, V., Rue, L., Woodgett, J. and Dedhar, S. (1998).Phosphoinositide-3-OH kinase-dependent regulation of glycogen synthase kinase 3 andprotein kinase B/AKT by the integrin-linked kinase. Proc. Natl. Acad. Sci. USA 95,11211-11216.

Feng, J., Park, J., Cron, P., Hess, D. and Hemmings, B. A. (2004). Identification of aPKB/Akt hydrophobic motif Ser-473 kinase as DNA-dependent protein kinase. J. Biol.Chem. 279, 41189-41196.

Förster, T. (1959). 10th Spiers Memorial Lecture. Transfer mechanisms of electronicexcitation. Discuss. Faraday Soc. 27, 7-17.

Forsythe, J. A., Jiang, B. H., Iyer, N. V., Agani, F., Leung, S. W., Koos, R. D. andSemenza, G. L. (1996). Activation of vascular endothelial growth factor genetranscription by hypoxia-inducible factor 1. Mol. Cell. Biol. 16, 4604-4613.

Harada, H., Hiraoka, M. and Kizaka-Kondoh, S. (2002). Antitumor effect of TAT-oxygen-dependent degradation-caspase-3 fusion protein specifically stabilized andactivated in hypoxic tumor cells. Cancer Res. 62, 2013-2018.

Hu, C. D., Chinenov, Y. and Kerppola, T. K. (2002). Visualization of interactions amongbZIP and Rel family proteins in living cells using bimolecular fluorescencecomplementation. Mol. Cell 9, 789-798.

Jiang, B. H., Jiang, G., Zheng, J. Z., Lu, Z., Hunter, T. and Vogt, P. K. (2001).Phosphatidylinositol 3-kinase signaling controls levels of hypoxia-inducible factor 1.Cell Growth Differ. 12, 363-369.

Kerppola, T. K. (2006a). Complementary methods for studies of protein interactions inliving cells. Nat. Methods 3, 969-971.

Kerppola, T. K. (2006b). Design and implementation of bimolecular fluorescencecomplementation (BiFC) assays for the visualization of protein interactions in livingcells. Nat. Protoc. 1, 1278-1286.

Legate, K. R., Montanez, E., Kudlacek, O. and Fassler, R. (2006). ILK, PINCH andparvin: the tIPP of integrin signalling. Nat. Rev. Mol. Cell. Biol. 7, 20-31.

Liu, Y., Cox, S. R., Morita, T. and Kourembanas, S. (1995). Hypoxia regulates vascularendothelial growth factor gene expression in endothelial cells. Identification of a 5�enhancer. Circ. Res. 77, 638-643.

Luker, K. E., Smith, M. C., Luker, G. D., Gammon, S. T., Piwnica-Worms, H. andPiwnica-Worms, D. (2004). Kinetics of regulated protein-protein interactions revealedwith firefly luciferase complementation imaging in cells and living animals. Proc. Natl.Acad. Sci. USA 101, 12288-12293.

Majumder, P. K., Febbo, P. G., Bikoff, R., Berger, R., Xue, Q., McMahon, L. M.,Manola, J., Brugarolas, J., McDonnell, T. J., Golub, T. R. et al. (2004). mTORinhibition reverses Akt-dependent prostate intraepithelial neoplasia through regulationof apoptotic and HIF-1-dependent pathways. Nat. Med. 10, 594-601.

Mongroo, P. S., Johnstone, C. N., Naruszewicz, I., Leung-Hagesteijn, C., Sung, R. K.,Carnio, L., Rustgi, A. K. and Hannigan, G. E. (2004). Beta-parvin inhibits integrin-linked kinase signaling and is downregulated in breast cancer. Oncogene 23, 8959-8970.

Paulmurugan, R., Umezawa, Y. and Gambhir, S. S. (2002). Noninvasive imaging ofprotein-protein interactions in living subjects by using reporter protein complementationand reconstitution strategies. Proc. Natl. Acad. Sci. USA 99, 15608-15613.

Persad, S., Attwell, S., Gray, V., Mawji, N., Deng, J. T., Leung, D., Yan, J., Sanghera,J., Walsh, M. P. and Dedhar, S. (2001). Regulation of protein kinase B/Akt-serine 473phosphorylation by integrin-linked kinase: critical roles for kinase activity and aminoacids arginine 211 and serine 343. J. Biol. Chem. 276, 27462-27469.

Rokudai, S., Fujita, N., Hashimoto, Y. and Tsuruo, T. (2000). Cleavage and inactivationof antiapoptotic Akt/PKB by caspases during apoptosis. J. Cell Physiol. 182, 290-296.

Sepulveda, J. L. and Wu, C. (2006). The parvins. Cell Mol. Life Sci. 63, 25-35.Tan, C., Cruet-Hennequart, S., Troussard, A., Fazli, L., Costello, P., Sutton, K.,

Wheeler, J., Gleave, M., Sanghera, J. and Dedhar, S. (2004). Regulation of tumorangiogenesis by integrin-linked kinase (ILK). Cancer Cell 5, 79-90.

Troussard, A. A., Mawji, N. M., Ong, C., Mui, A., St -Arnaud, R. and Dedhar, S.(2003). Conditional knock-out of integrin-linked kinase demonstrates an essential rolein protein kinase B/Akt activation. J. Biol. Chem. 278, 22374-22378.

Williams, M. R., Arthur, J. S., Balendran, A., van der Kaay, J., Poli, V., Cohen, P. andAlessi, D. R. (2000). The role of 3-phosphoinositide-dependent protein kinase 1 inactivating AGC kinases defined in embryonic stem cells. Curr. Biol. 10, 439-448.

Yamaji, S., Suzuki, A., Sugiyama, Y., Koide, Y., Yoshida, M., Kanamori, H., Mohri, H.,Ohno, S. and Ishigatsubo, Y. (2001). A novel integrin-linked kinase-binding protein, affixin,is involved in the early stage of cell-substrate interaction. J. Cell Biol. 153, 1251-1264.

Yamaji, S., Suzuki, A., Kanamori, H., Mishima, W., Yoshimi, R., Takasaki, H.,Takabayashi, M., Fujimaki, K., Fujisawa, S., Ohno, S. et al. (2004). Affixin interactswith alpha-actinin and mediates integrin signaling for reorganization of F-actin inducedby initial cell-substrate interaction. J. Cell Biol. 165, 539-551.

Jour

nal o

f Cel

l Sci

ence