Protein Expression and Purification 78 (2011) 174–180

Contents lists available at ScienceDirect

Protein Expression and Purification

journal homepage: www.elsevier .com/ locate /yprep

Expression in Escherichia coli and purification of human recombinant connexin-43,a four-pass transmembrane protein

Sédami Gnidehou a,b,1,⇑, Pascale Gerbaud a,b,c,1, Guillaume Ducarme a, Fatima Ferreira a,b,c, Josette Badet a,b,André Malassiné a,b,c, Danièle Evain-Brion a,b,c, Jean-Louis Frendo a,b,d

a INSERM, U767, Paris F-75006, Franceb Université Paris Descartes, Paris F-75006, Francec PremUp, Paris F-75006, Franced CNRS, Paris F-75006, France

a r t i c l e i n f o

Article history:Received 19 January 2011and in revised form 15 April 2011Available online 4 May 2011

Keywords:Connexin-43Expression vectorPlasmidRecombinant proteinTrophoblastZona occludens-1

1046-5928/$ - see front matter � 2011 Elsevier Inc. Adoi:10.1016/j.pep.2011.04.018

Abbreviations: Cx43, connexin-43; cDNA, compacid; CDNB, 1-chloro-2,4-dinitrobenzene; GJIC, gapmunication; GST, glutathione S-transferase; LB, Luriabination; ZO-1, Zona Occludens-1.⇑ Corresponding author. Address: INSERM U767, F

ceutiques et Biologiques, 4 Avenue de l’Observatoire, 744 07 39 92.

E-mail address: gcarine@yahoo.com (S. Gnidehou)1 These two authors contributed equally to the work

a b s t r a c t

We have recently shown, using a well-defined in vitro model, that connexin 43 (Cx43) is directly involvedin human cytotrophoblastic cell fusion into a multinucleated syncytiotrophoblast. Cx43 appears to inter-act with partner proteins within a fusogenic complex, in a multi factorial and dynamic process. This fus-ogenic complex remains to be characterized and constituent proteins need to be identified. In order toidentify proteins interacting with the entire Cx43 molecule (extracellular, transmembrane and intracel-lular domains), we produced and purified full-length recombinant Cx43 fused to glutathione S-transfer-ase (GST-Cx43) and used it as ‘‘bait’’ in GST pull-down experiments. Cx43 cDNA was first cloned into thepDEST15 vector in order to construct a GST-fusion protein, using the Gateway system. The fusion proteinGST-Cx43 was then expressed in Escherichia coli strain BL21-AI™ and purified by glutathione-affinitychromatography. The purified fusion protein exhibited the expected size of 70 kDa on SDS–PAGE, westernblot and GST activity. A GST pull-down assay was used to show the capacity of the full-length recombi-nant protein to interact with known partners. Our results suggest that this method has the capacity toproduce sufficient full-length recombinant protein for investigations aimed at identifying Cx43 partnerproteins.

� 2011 Elsevier Inc. All rights reserved.

Introduction

Cell fusion plays an essential role in several major biologicalprocesses, including fertilization, embryonic and fetal develop-ment, cell differentiation, tissue repair and possibly, tumor devel-opment [1–3]. Few human cell types can fuse together anddifferentiate into multinucleated cells. Myoblasts fuse to form mul-tinucleated skeletal muscle fibers (myotubes) [4–6]; cells of mono-cytic origin fuse to form osteoclasts, which participate in bonesculpturing and remodeling [7]; and cytotrophoblasts fuse to forma multinucleated syncytiotrophoblast [8,9].

ll rights reserved.

lementary deoxyribonucleicjunctional intracellular com-–Bertani; LR, lambda recom-

aculté des Sciences Pharma-5270 Paris, France. Fax: +33 1

..

Little is known of the molecular mechanisms of cell–cell fusionprocess, despite the development of yeast, nematode, insect andmammalian models [10]. Primary cultured human cytotropho-blasts are a useful model for investigating this issue, as a key featureof human placental development is the fusion of cytotrophoblasticcells to form a multinucleated syncytiotrophoblast, a process thatcan be reproduced in vitro [11]. Purified cytotrophoblasts isolatedfrom human placentas aggregate and then fuse, forming a multinu-cleated syncytiotrophoblast. We have recently demonstrated, usingthis model, the direct involvement of human endogenous retroviralenvelope Herv-W (syncytin 1) [9], gap junctional intracellular com-munication (GJIC) [12], zona occludens-1 (ZO-1) [13] and connexin43 (Cx43) [8] proteins in cytotrophoblastic cell fusion.

Human Cx43 (also called Gap junction alpha-1 protein) be-longs to the connexin protein family. The Cx43 polypeptide chainthreads through the lipid bilayer four times, forming four alpha-helical transmembrane domains, two extracellular loops and acytoplasmic loop. The amino and carboxy termini are located inthe cytoplasm [14,15]. The extracellular loops are responsiblefor interactions between connexons in apposed cell membranes.Each connexon is composed of six connexin subunits, forming a

S. Gnidehou et al. / Protein Expression and Purification 78 (2011) 174–180 175

hemi-channel located in the plasma membrane. Two hemi-chan-nels from different cells dock with each other to form a contigu-ous gap junction channel, allowing direct communicationbetween neighboring cells. Gap junctions allow the passage be-tween cells of small molecules such as metabolites (glucose andamino acids), ions (K+ and Ca2+) and second messengers (cAMP,cGMP and IP3) [16]. They play an essential role in placental func-tions and development [17].

There is accumulating evidence that Cx43 interacts withseveral proteins, including cytoskeletal proteins and anchoringproteins such as E-cadherin [18], caveolin-1 [19] and ZO-1[13,20–22]. A recent study using nuclear magnetic resonanceshowed that Cx43 interaction with ZO-1 occurs through the 20C-terminal amino acids of Cx43 and the second PDZ domain ofZO-1 [20,23]. These interactions may regulate channel assemblyand Cx43 stability at the plasma membrane. As we have previ-ously shown that Cx43 is directly involved in trophoblastic cellfusion [8], we used this protein as ‘‘bait’’ to identify and charac-terize other proteins composing the fusogenic complex. Toidentify proteins interacting with the extracellular and/or intra-cellular domains of Cx43, we used glutathione S-transferase(GST) pull-down experiments with the full-length Cx43 fusedto GST. Most previous studies of Cx43 interactions have usedco-immunoprecipitation, yeast two-hybrid system or pull-downassays, using peptides corresponding to known Cx43 sequences(mainly the intracytoplasmic C-terminal tail) fused to GST[20,21,24,25]. For our purpose, we needed to produce and purifya full-length recombinant protein GST-Cx43. In order to expressand purify such a protein, it was necessary to develop an efficientand convenient protocol. Here we describe the first successful at-tempt to express in Escherichia coli and purify full-length humanrecombinant Cx43, using the GST fusion system.

Materials and methods

Reagents

E. coli TOP10 competent cells (used for cloning and plasmidamplification) and E. coli strain BL21-AI™ (used to express the fu-sion protein) were both from Invitrogen (Carlsbad, CA, USA).PDEST15, pENTR/SD/D-TOPO, the LR clonase II enzyme reactionmixture, arabinose, and acrylamide minigels were from Invitrogen.

Cloning and construction of the recombinant plasmid

Gateway Technology (Invitrogen) was used to construct, inE. coli, the expression system consisting of the Cx43 gene fused tothe GST gene. This technology takes advantage of the site-specificrecombination properties of bacteriophage lambda (LR) to fuseCx43 to GST both rapidly and efficiently. First, the Cx43 coding re-gion was cloned in the pENTR/SD/D-TOPO plasmid to generate anentry clone. Then, recombination reaction between attL and attRsites (LR recombination reaction) was achieved between the entryclone and the destination vector pDEST15 (an N-terminal GST fu-sion vector containing the T7 promoter) to generate the expressionconstruct pDEST15-GST-Cx43. Briefly, the coding region of Cx43was amplified from a plasmid containing the complete sequenceof human connexin 43 (kindly provided by Dr. M. Mesnil) by poly-merase chain reaction using Pfx DNA Platinum polymerase and twoprimers (50-CACCATGGGTGACTGGAGCGC-30 and 50-stop codonCTAGATCTCCAGGTCATCAGGCCG). Amplification was run for 30 cy-cles consisting of denaturation at 94 �C for 30 s, primer annealing at60 �C for 30 s and primer extension at 72 �C for 1 min. The amplifi-cation product was purified from agarose gel by using the GelExtraction and Purification kit. The purified cDNA bearing CACC at

the 50-end was cloned into the pENTR/SD/D-TOPO plasmid to createthe Cx43-pENTR/SD/D-TOPO entry clone. Then, the gene wastransferred into the destination vector pDEST15 through an LRrecombination reaction to generate the expression clonepDEST15-GST-Cx43, following the manufacturer’s instructions.Cx43-pENTR/SD/D-TOPO and pDEST15 (150 ng each) were addedto a 10-lL LR clonase II enzyme reaction mixture and incubatedfor 1 h at 25 �C. The reaction was stopped by adding proteinaseK. E. coli TOP10 competent cells were transformed with 1 lL ofthe recombinant plasmid pDEST15-GST-Cx43 and selected inLuria–Bertani (LB) broth containing 100 lg/mL ampicillin. Positiveclones were selected by PCR and sequenced by Cogenix (France).Both strands of DNA fragments were sequenced using M13 reverseand forward primers. The Gateway cloning method was also used togenerate a recombinant GST for use as a positive control. ThepDEST15 plasmid containing the GST gene was subjected to BamH1digestion in order to delete the Ccdb gene (the product of this gene islethal for bacterial cells) and to introduce a stop codon after the GSTsequence. The purified linearized plasmid was ligated overnight byusing 1 U of T4 DNA ligase (New England Biolabs). The ligated vectorwas then transformed into E. coli TOP10. Clones with mutated Ccdbare ampicillin- and chloramphenicol-resistant. Positive clones weresequenced. The pDEST15-GST and pDEST15-GST-Cx43 vectors weretransformed into E. coli BL21-AI™ to produce recombinant GST andGST-Cx43 fusion proteins, respectively.

Expression of GST-Cx43 fusion protein

E. coli BL21-AI™ carries a chromosomally inserted cassette con-taining the T7 RNA polymerase gene at the araB locus of thearaBAD operon, allowing T7 RNA polymerase expression to be reg-ulated by the araBAD promoter. The araB gene is deleted in thisstrain. E. coli BL21-AI™ cells transformed with the pDEST15-GSTvector or the pDEST15-GST/Cx43 plasmid were cultured at 30 �Cor 37 �C, then treated with arabinose to induce protein production.To determine the time course of GST-Cx43 protein expression, ali-quots removed just before and 1, 2, 3 and 4 h after arabinose induc-tion were lysed in Laemmli buffer (62.5 mM Tris–HCl, pH 6.8, 2%(w/v) SDS, 5% (v/v) b-mercaptoethanol, 10% (v/v) glycerol) for10 min at 70 �C and analyzed by 4–12% SDS–PAGE as described be-low. The fusion protein was expressed on a large scale and purified.Briefly, cells were grown overnight (16 h) in 50 mL of LB broth con-taining 100 lg/mL carbenicillin with gentle shaking at 37 �C. Thefollowing morning, 10 mL of culture was inoculated into 1 litre ofLB broth and incubated at 37 �C until the culture reached themid-log phase (OD600 = 0.4). Adding arabinose at a final concentra-tion of 0.2% (w/v) then induced protein expression. The cells werecultured at 30 �C for a further 1 h and then harvested by centrifu-gation at 15,000g for 15 min at 4 �C. The pellet was washed in coldphosphate-buffered saline (PBS, 140 mM NaCl, 2.7 mM KCl, 10 mMNa2HPO4, 1.8 mM KH2PO4, pH 7.3) containing protease inhibitors(Cocktail Set 1, Calbiochem).

Affinity purification of GST-Cx43 fusion protein

The cell pellet (10 g wet weight) was resuspended in 40 mL ofcold PBS buffer containing protease inhibitors. The cell suspensionwas sonicated on ice using Branson Sonifier 450 equipped with a6-mm-diameter microtip (2 � 3 min in the pulsed mode, duty cycle15%, output control at 5). The sonicated material was centrifuged at15,000g for 15 min at 4 �C. The supernatant and pellet wereanalyzed by 4–12% SDS–PAGE and immunoblotted. The proteinconcentration was determined with the Nanodrop spectrophotom-eter (Labtech, Palaiseau, France). The GST-Cx43 fusion protein wasaffinity-purified by using Glutathione (GSH)–Sepharose beads

Fig. 1. Construction of the pDEST15-Cx43 expression vector. (A) Human Cx43 cDNA was amplified and cloned in pENTR/SD/D-TOPO to create the entry clone. (B) Homologousrecombination between the attL sites of the entry clone vector and the attR sites of the pDEST15 vector was done, resulting in the fusion of Cx43 downstream of GST. The newvector pDEST15-GST-Cx43 was used to produce Cx43 recombinant protein.

176 S. Gnidehou et al. / Protein Expression and Purification 78 (2011) 174–180

(G4551, Sigma) and a batch method. Briefly, the supernatant(20 mL) containing the fusion protein was transferred to 40 mg ofGSH-Sepharose beads suspension that had been pre-equilibratedthree times with PBS, then incubated with gentle agitation for90 min at room temperature. After centrifugation at 500g for5 min, the sedimented matrix was washed three times with 5 mLof PBS. The GST-Cx43 fusion protein was then eluted with 50 mMTris–HCl, pH 7.4 containing 10 mM reduced GSH and 300 mM NaCl.In preliminary experiments various NaCl concentrations (150, 300and 500 mM) were tested. The protein was collected in a volumeof 0.5-mL. The eluates were pooled, desalted and concentratedusing Centripep-10 (Amicon Ultra 10 K, Millipore), following themanufacturer’s recommendations, and stored at �80 �C. The yieldof eluted GST or GST-Cx43 was determined with Nanodrop spectro-photometer (Molar extinction coefficient: 87,000 M�1 cm�1) andGST activity assays. To evaluate purity, eluted proteins were runon 4–12% SDS–PAGE, followed by Coomassie brilliant blue R-250staining.

SDS–PAGE and immunoblotting

Protein (20–50 lg) was solubilized in Laemmli buffer, heatedfor 10 min at 70 �C and subjected to SDS–PAGE using 4–12% or3–8% acrylamide slab minigels (Invitrogen, Carlsbad, CA). Proteinwas electrotransferred to 0.2-lm Protan BA 83 nitrocellulosemembrane (Schleicher and Schuell, Dassel, Germany) using aiBlot™ Gel Transfer System (Invitrogen). The blot was saturatedwith PBS supplemented with 5% (w/v) nonfat dry milk and 0.1%(v/v) Tween� 20 for 1 h at room temperature and then incubatedwith either a 1/10,000 dilution of a polyclonal antibody againstCx43 (C6219, Sigma) or a 1/5000 dilution of a polyclonalantibody against GST (G7781, Sigma). After five washes withPBS for 3 min, immune complexes were detected with a horserad-ish peroxidase-coupled anti-rabbit IgG antibody (1/30,000; 111–035-046, Jackson ImmunoResearch, Baltimore, MD) for 1 h,followed by phototope western detection system at room temper-ature (Supersignal, Pierce, Rockford, IL). The membrane was then

exposed to X-Omat film (Eastman Kodak, Rochester, NY) for sig-nal detection.

GST activity assay

The active GST-Cx43 fusion protein was detected by evaluatingthe GST activity of the total bacterial lysate, soluble fraction andpurified protein, using the GST substrate 1-chloro-2,4-dinitroben-zene (CDNB) [26,27]. GST mediates a reaction between CDNB andglutathione, producing a conjugate that can be assayed by mea-suring absorbance at 340 nm. Briefly, 880 lL of distilled water,100 lL of 1 M potassium phosphate pH 6.5, and 10 lL of100 mM CDNB in ethanol were mixed, and 10 lL of 100 mM re-duced glutathione was added (CDNB reaction solution). Then 5–10 lL of sample or an equal volume of reaction buffer (blank)was added to 100 lL of the CDNB reaction solution. Purified re-combinant GST was used as a positive control. Absorbance wasmonitored at 340 nm for 5 min with a microplate spectrophotom-eter. The absorbance of the blank was subtracted from experi-mental values carried out in triplicate.

GST pull-down assay and immunoblot

The GST pull-down assay was done according to the manufac-turer’s recommendations. Briefly, cytotrophoblastic cells werelysed in binding buffer (MagneGST Protein Purification System,Promega V 8600) containing protease inhibitors (protease inhibitorcocktail, Set I-539131, Calbiochem) for 30 min at 4 �C, then centri-fuged at 12,000g for 10 min at 4 �C. Purified GST-Cx43 or GST alone(100 lg) coupled to 100 lL of pre-equilibrated MagneGST particleswas incubated with clarified cytotrophoblast extracts in bindingbuffer. The mixture was incubated on a rotating platform for 1 hat room temperature. The suspension was placed in a magneticstand to capture the MagneGST particles. Pelleted beads werewashed three times with binding buffer and the remaining proteincomplexes were eluted with LDS 1� sample buffer (NP0008, Invit-rogen) then subjected to SDS–PAGE.

S. Gnidehou et al. / Protein Expression and Purification 78 (2011) 174–180 177

Placental trophoblast cell culture

These studies were approved by our local ethics committee(Comité de Protection des Personnes Île-de-France III) and thedonors gave their written consent. Placentas were obtained imme-diately after elective Caesarian section from healthy mothers withuncomplicated pregnancies delivered at 35–39 weeks of amenor-rhea. Cytotrophoblasts were isolated as previously described[28]. After sequential trypsin/DNAse I digestion followed by Percollgradient centrifugation [8], cytotrophoblasts were diluted to a finaldensity of 2.7 � 106 cells in 3 mL of Dulbecco’s modified Eagle’smedium containing 10% fetal calf serum. Cells were plated in 60-mm plastic dishes (TPP, Trasadingen, Switzerland) and incubatedat 37 �C in 5% CO2-air. About 95–98% of the cells stained positivelyfor cytokeratin 7, confirming their trophoblastic nature.

Statistical analysis

The StatView F-4.5 software package (Abacus Concepts, Inc., CA)was used for all analyses. Values are reported as means ± SE.

Fig. 2. Expression of GST-Cx43 fusion protein in E. coli BL21-AI™ SDS–PAGE and western(A and C) and 37 �C (B and D). Total protein of BL21-AI™ containing pDEST15-GST-Cxseparated by 4–12% SDS–PAGE (upper), transferred to a nitrocellulose membrane and proGST-Cx43. (C) and (D) represent lysates of E. coli BL21-AI™ transformed with the pDEST1largest amount of fusion protein was produced after 1 or 2 h of induction at 30 �C, wheduring induction. GSTrec corresponds to commercial GST used as a positive control.

Significant differences (p < 0.05) were identified by using analysisof variance (ANOVA).

Results

Construction of the GST-Cx43 fusion expression vector

The constructed plasmid (pDEST15-GST-Cx43) used to expressthe recombinant protein is shown in Fig. 1. To generate this expres-sion vector, we first PCR-amplified Cx43 cDNA and then introducedthe resulting fragment into the pENTR/SD/D-TOPO plasmid, be-tween recombination sequences attL1 and attL2, to create theCx43-pENTR/SD/D-TOPO entry clone (Fig. 1A). These recombina-tion sequences allow recombinational cloning of the gene of inter-est from the entry construct within the attR1 and attR2 sequencespresent in the destination vector pDEST15, containing the GSTgene. By homologous recombination, the Cx43 gene was inserteddownstream of GST-encoding DNA in the same open reading frame(Fig. 1B). The orientation and sequence were verified by DNAsequencing (not shown). Cx43 was successfully cloned with GSTfused to its N-terminal extremity.

blot analysis of E. coli BL21-AI™ lysates after induction with 0.2% arabinose at 30 �C43, after induction with 0.2% arabinose or PBS (control) for 0, 1, 2, 3 and 4 h wasbed with anti-Cx43 (lower). Molecular weight marker in lane 1. The arrow indicates5-GST vector treated in the same conditions but probed with anti-GST (lower). The

reas, whatever the temperature, the amount of GST recombinant protein increased

Fig. 3. Purification of the GST-Cx43 fusion protein expressed in E. coli BL21-AI™. (A): SDS–PAGE and western blot analysis of the soluble fraction (SF) and pellet (P) obtainedfrom lysed bacterial cells (TF) after induction with 0.2% arabinose at 30 �C for 1 h. Right panel: the arrow indicates GST-Cx43. (B): SDS–PAGE analysis of GST-Cx43 fusionprotein purified by affinity chromatography. GST-Cx43 was eluted from the GSH-Sepharose beads by using various concentrations of NaCl in the presence of 10 mM reducedGSH. (C): Purified, desalted and concentrated GST-Cx43 recombinant protein was checked by SDS–PAGE and western blot with anti-Cx43. The arrow indicates GST-Cx43. (SF):soluble fraction, (E): eluate.

Table 1Recombinant GST-Cx43 purification. Activity and yield of purified GST-Cx43 fusionprotein. GST activity was determined in triplicate in the soluble and eluted fractionsby using the GST substrate CDNB. ⁄⁄p < 0. 001. All values are the mean ± SD of 12determinations.

Soluble fraction Elution fraction

Volume (ml) 160 0.5Protein concentration (mg/ml) 4.9 ± 0.5 5.8 ± 0.4Protein quantity (mg) 781 ± 70 2.9 ± 0.15Activity (U) 1.2 ± 0.1 0.096 ± 0.002Specific activity (mU/mg) 1.40 ± 0.15 27 ± 1Yield (%) 100 7.2 ± 0.7Purification factor 1 19.7 ± 3.5

178 S. Gnidehou et al. / Protein Expression and Purification 78 (2011) 174–180

Expression of GST-Cx43 fusion protein in E. coli BL21-AI

In order to produce the GST-Cx43 fusion protein, the plasmidpDEST15-GST-Cx43 was expressed in E. coli BL21-AI™. Several cul-ture parameters, including the induction conditions (i.e. inductiontiming and arabinose concentration), agitation speed and temper-ature were found to critically affect the culture performance. Wefirst determined the temperature (30 or 37 �C) and period (1, 2, 3or 4 h) required for optimal induction by arabinose. UsingSDS–PAGE gels stained with Coomassie blue and western blot anal-ysis with an anti-Cx43 rabbit antibody, we found that, after induc-tion with 0.2% arabinose, the bacteria produced a protein that wasabsent in non-inducing conditions (Fig. 2A, 2B). As expected, onSDS–PAGE, this protein had a molecular weight of approximately70 kDa, corresponding to the fusion of GST (27 kDa) and Cx43(43 kDa). The expression level of GST-Cx43 declined with theinduction time at 30 �C (Fig. 2A) and at 37 �C (Fig. 2B). The largestamount of fusion protein was produced after 1 h or 2 h of induc-tion, the yield reaching a plateau and declining thereafter. Thismay be because E. coli proteases target the recombinant proteinwhen over-expressed in the prokaryotic system, resulting in a re-duced protein yield. Production efficiency was better after induc-tion at 30 �C than at 37 �C. The positive control was the pDEST15vector modified to produce GST protein alone (see Section Materi-als and Methods). E. coli BL21-AI™ transformed with this modifiedpDEST15 vector produced a protein of 27 kDa (GST protein) afterarabinose induction, as shown by SDS–PAGE and western blot(Fig. 2C and D). As shown in Fig. 2, E. coli BL21-AI™ transformedwith GST-Cx43- or GST-pDEST15 plasmid expressed recombinant

proteins with respective molecular weight of 70 and 27 kDa, as ex-pected. The optimal conditions for GST-Cx43 fusion protein expres-sion were 30 �C and 1 h of induction with 0.2% arabinose.

Affinity purification of GST-Cx43

We first checked that the recombinant protein was expressed ina soluble form. Bacterial lysates obtained after growth at 30 �C andinduction with 0.2% arabinose for 1 h were centrifuged. Part ofGST-Cx43 fusion protein was detected in the soluble fraction(Fig. 3A). GST-Cx43 was then purified from the soluble fractionby affinity chromatography on GSH-Sepharose beads. PurifiedGST-Cx43 contained a single major product of the expected size(Fig. 3B). Optimal elution was obtained with elution buffer

Fig. 4. Specific interaction between recombinant GST-Cx43 and human ZO-1 in a GST pull-down assay. Immobilized purified GST or GST-Cx43 was incubated withcytotrophoblast lysate. (A) SDS–PAGE analysis of purified GST (1), GST-Cx43 (2) and the retained proteins (3 and 4, respectively). (B) Immunoblot probed with anti-ZO1antibodies: human cytotrophoblast lysate (a), retained proteins after incubation with GST (b) or GST-Cx43 (c).

S. Gnidehou et al. / Protein Expression and Purification 78 (2011) 174–180 179

containing 10 mM reduced GSH and 300 mM NaCl (Fig. 3B). Thespecificity of the recombinant protein was further analyzed bywestern blot with a rabbit anti-Cx43 antibody (Fig. 3C). An intense70-kDa band was recognized, confirming that GST-Cx43 proteinhad been successfully purified and enriched. The yield of purifiedGST-Cx43 was about 3 mg per litre of culture.

Glutathione S-transferase activity of GST-Cx43

We assayed GST activity to estimate the purification yield. Theactivity of the GST-Cx43 fusion protein was determined by usingthe artificial GST substrate CDNB. GST activity was identified inthe eluted fraction (27 ± 1 mU.mg�1 versus 1.40 ± 0.15 mU mg�1

in the soluble fraction; Table 1), the yield and purification factorare 7% and 20% respectively (Table 1). Thus, the GST domain ofthe recombinant protein was not denatured and seemed to be cor-rectly folded.

In vitro interaction between GST-Cx43 fusion protein and ZO-1

To examine GST-Cx43 interaction with its known partner pro-teins, we used a GST pull-down binding assay. GST-Cx43 coupledto GSH-Sepharose beads was incubated with cytotrophoblast ly-sates. After washing steps, proteins associated with GST-Cx43 wereresolved by SDS–PAGE (Fig. 4A). As it is known to interact withCx43, we sought ZO-1 among the eluted proteins. Immunoblotswith anti-ZO-1 revealed an intense band at 220 kDa, correspondingto the molecular weight of ZO-1 (Fig. 4B lane c). To determinewhether this interaction was specific to a Cx43 domain, we incu-bated the same cytotrophoblast lysate with GST alone coupled tothe beads (Fig. 4B lane b), and observed no corresponding band.This GST pull-down experiment suggested that the recombinantCx43 could interact with its partner proteins.

Discussion

We report the first successful attempt to express in E. coli andpurify a full-length human recombinant Cx43 protein using theGST fusion system. E. coli is the expression host of choice, givenits well-studied physiology and relatively simple genetics, theavailability of advanced genetic tools, relatively rapid growth, easeof handling in a standard laboratory, low cost, and the feasibility ofboth expression screening and protein production. However, mostpost-translational modifications are absent and the recombinantproteins are frequently expressed as insoluble aggregates of

folding intermediates known as inclusion bodies [29,30]. One com-mon successful strategy used to enhance solubility is to fuse theprotein of interest with a highly soluble tag protein such as GST[31], maltose binding protein (MBP) or thioredoxin (Trx) [32–34].We chose to use GST, based on the results of a study ranking tagsin terms of soluble expression and purification yields: GST andMBP were the most efficient tags when comparing the expressionof 32 human proteins with molecular weights ranging from 17 to110 kDa [33]. Another strategy to enhance soluble expression andyield is to truncate Cx43 into separate domains [20,21,24,25].However, this approach would have required us to produce and pur-ify 8 different constructs spanning the two extracellular domains,four transmembrane domains and two intracellular domains.

We thus cloned human Cx43 cDNA in pDEST15, an E. coliexpression vector containing the GST coding region of Schistosomajaponicum, thereby generating the recombinant plasmid pDEST15-GST-Cx43. E. coli BL21-AI™ transformed with plasmid pDEST15-GST-Cx43 and induced with arabinose expressed a protein withthe expected molecular mass (70 kDa) corresponding to the GST-Cx43 fusion protein. The identity of the recombinant protein wasconfirmed by western blot and CDNB assay. Although the GSTexpression system is designed for inducible, high-level intracellu-lar expression of soluble proteins [31], many factors can influencethe yield of the fusion protein. In our study, a relative large amountof GST-Cx43 fusion protein was present in its insoluble form. Ini-tially, we attempted to extract the recombinant fusion protein inthe presence of denaturing agent. However the refolding of theprotein was not ensured as shown by the loss of its GST activity.In order to reduce the formation of insoluble aggregates, we unsuc-cessfully tried a lower concentration of arabinose, a shorter expres-sion period, a lower expression temperature and increased cultureaeration. One hour of induction with 0.2% arabinose at 30 �C pro-vided the most efficient production of recombinant GST-CX43 pro-tein, in keeping with the increased solubility of recombinantproteins at lower temperatures [30].

GSH-affinity chromatography is a relatively simple and efficientmethod for purifying fusion proteins. We recovered a major bandon SDS–PAGE and western-blot of a size (70 kDa) correspondingto the GST-Cx43 fusion protein. About 3 mg of purified fusion pro-tein was obtained from one litre of bacterial culture. We are awareof no other reports on the expression and purification of full-length4-transmembrane-domain proteins, but this yield is comparable tothat reported for other purified 70-kDa fusion proteins [35–38].Our method proved highly reproducible: we repeated the expres-sion and purification steps and observed only very slight variationsin yield and GST activity.

180 S. Gnidehou et al. / Protein Expression and Purification 78 (2011) 174–180

The ability of GST-Cx43 fusion protein to bind on the GSH-Sepharose beads was essential to perform GST pull-down experi-ments. GST activity of the GST-Cx43 fusion protein was found tobe concentration-dependent. The presence of an N-terminal tagcould affect the conformation of the Cx43 protein and its interac-tion with partner proteins. An intact Cx37 N-terminus has beenshown to be required for function but not gap junction formation[39]. Nevertheless, this work presents that ZO-1, a well knownpartner of Cx43 [20–22], binds to the purified fusion protein, indi-cating that the recombinant Cx43 protein might interact withother partner proteins. Work in progress confirms the potentialof the recombinant Cx43 protein.

In conclusion, we report a simple protocol for expression andpurification, in a reproducible manner, of a recombinant full-lengthCx43 protein. This fusion protein will now be used as a tool to iden-tify partners interacting with Cx43 and thereby to unravel cell–cellfusion processes, mainly in trophoblast cell fusion, a limiting stepin human villous trophoblast differentiation and an essential eventfor efficient placental functions.

Footnotes

This work was supported by Caisse d’Assurance Maladie desProfessions Libérales Province and by Institut National de la Santéet de la Recherche Médicale (Projet Avenir grant to JL Frendo).

Acknowledgments

We thank Dr. Fanny Lewin for her support, the staff of the SaintVincent de Paul Obstetrics Department for providing us with pla-centas, Bernard Le Bonniec for his advice and invaluable discus-sions, and David Young for editorial assistance. S. Gnidehou wassupported by a fellowship from Institut National de la Santé etde la Recherche Médicale (Projet Avenir grant to JL Frendo).

References

[1] P. Sutovsky, Sperm-egg adhesion and fusion in mammals, Expert Rev. Mol.Med. 1 (2009). 11:e11.

[2] X. Lu, Y. Kang, Cell fusion as a hidden force in tumor progression, Cancer Res.69 (2009) 8536–8539.

[3] L.I. Larsson, B. Bjerregaard, J.F. Talts, Cell fusions in mammals, Histochem. CellBiol. 129 (2008) 551–561.

[4] K.M. Jansen, G.K. Pavlath, Molecular control of mammalian myoblast fusion,Methods Mol. Biol. 475 (2008) 115–133.

[5] R.M. Mege, D. Goudou, C. Giaume, M. Nicolet, F. Rieger, Is intercellularcommunication via gap junctions required for myoblast fusion?, Cell AdhesCommun. 2 (1994) 329–343.

[6] M.J. Wakelam, The fusion of myoblasts, Biochem. J. 228 (1985) 1–12.[7] A. Zambonin Zallone, A. Teti, M.V. Primavera, Monocytes from circulating

blood fuse in vitro with purified osteoclasts in primary culture, J. Cell Sci. 66(1984) 335–342.

[8] J.-L. Frendo, L. Cronier, G. Bertin, J. Guibourdenche, M. Vidaud, D. Evain-Brion,A. Malassine, Involvement of connexin 43 in human trophoblast cell fusion anddifferentiation, J. Cell Sci. 116 (2003) 3413–3421.

[9] J.-L. Frendo, D. Olivier, V. Cheynet, J.-L. Blond, M. Vidaud, M. Rabreau, D. Evain-Brion, F. Mallet, Direct involvement of HERV-W Env glycoprotein in humantrophoblast cell fusion and differentiation, Mol. Cell. Biol. 23 (2003) 3566–3574.

[10] E.H. Chen, E.N. Olson, Unveiling the mechanisms of cell–cell fusion, Science308 (2005) 369–373.

[11] H.J. Kliman, J.E. Nestler, E. Sermasi, J.M. Sanger, J.F. Strauss III, Purification,characterization, and in vitro differentiation of cytotrophoblasts from humanterm placentae, Endocrinology 118 (1986) 1567–1582.

[12] L. Cronier, J.-L. Frendo, N. Defamie, G. Pidoux, G. Bertin, J. Guibourdenche, G.Pointis, A. Malassine, Requirement of gap junctional intercellularcommunication for human villous trophoblast differentiation, Biol. Reprod.69 (2003) 1472–1480.

[13] G. Pidoux, P. Gerbaud, S. Gnidehou, M. Grynberg, G. Geneau, J. Guibourdenche,D. Carette, L. Cronier, D. Evain-Brion, A. Malassiné, J.-L. Frendo, ZO-1 is

involved in trophoblastic cell differentiation in human placenta, Am. J. Physiol.Cell Physiol. 298 (2010) C1517–1526.

[14] E.C. Beyer, D.L. Paul, D.A. Goodenough, Connexin43: a protein from rat hearthomologous to a gap junction protein from liver, J. Cell Biol. 105 (1987) 2621–2629.

[15] C.K. Manjunath, G.E. Goings, E. Page, Human cardiac gap junctions: isolation,ultrastructure, and protein composition, J. Mol. Cell. Cardiol. 19 (1987) 131–134.

[16] N.M. Kumar, N.B. Gilula, The gap junction communication channel, Cell 84(1996) 381–388.

[17] A. Malassine, L. Cronier, Involvement of gap junctions in placental functionsand development, Biochim. Biophys. Acta 1719 (2005) 117–124.

[18] K. Fujimoto, A. Nagafuchi, S. Tsukita, A. Kuraoka, A. Ohokuma, Y. Shibata,Dynamics of connexins, E-cadherin and alpha-catenin on cell membranesduring gap junction formation, J. Cell Sci. 110 (1997) 311–322.

[19] A.L. Schubert, W. Schubert, D.C. Spray, M.P. Lisanti, Connexin family memberstarget to lipid raft domains and interact with caveolin-1, Biochemistry 41(2002) 5754–5764.

[20] B.N. Giepmans, W.H. Moolenaar, The gap junction protein connexin43interacts with the second PDZ domain of the zona occludens-1 protein, Curr.Biol. 8 (1998) 931–934.

[21] T. Toyofuku, M. Yabuki, K. Otsu, T. Kuzuya, M. Hori, M. Tada, Direct associationof the gap junction protein connexin-43 with ZO-1 in cardiac myocytes, J. Biol.Chem. 273 (1998) 12725–12731.

[22] J.C. Wu, R.Y. Tsai, T.H. Chung, Role of catenins in the development of gapjunctions in rat cardiomyocytes, J. Cell. Biochem. 88 (2003) 823–835.

[23] P.L. Sorgen, H.S. Duffy, P. Sahoo, W. Coombs, M. Delmar, D.C. Spray, Structuralchanges in the carboxyl terminus of the gap junction protein connexin43indicates signaling between binding domains for c-Src and zonula occludens-1, J. Biol. Chem. 279 (2004) 54695–54701.

[24] B.N. Giepmans, I. Verlaan, T. Hengeveld, H. Janssen, J. Calafat, M.M. Falk, W.H.Moolenaar, Gap junction protein connexin-43 interacts directly withmicrotubules, Curr. Biol. 11 (2001) 1364–1368.

[25] C. Jin, A.F. Lau, K.D. Martyn, Identification of connexin-interacting proteins:application of the yeast two-hybrid screen, Methods 20 (2000) 219–231.

[26] U.H. Danielson, B. Mannervik, Paradoxical inhibition of rat glutathionetransferase 4–4 by indomethacin explained by substrate-inhibitor-enzymecomplexes in a random-order sequential mechanism, Biochem. J. 250 (1988)705–711.

[27] W.H. Habig, M.J. Pabst, W.B. Jakoby, Glutathione S-transferases. The firstenzymatic step in mercapturic acid formation, J. Biol. Chem. 249 (1974) 7130–7139.

[28] E. Alsat, J. Haziza, D. Evain-Brion, Increase in epidermal growth factor receptorand its messenger ribonucleic acid levels with differentiation of humantrophoblast cells in culture, J. Cell. Physiol. 154 (1993) 122–128.

[29] D.C. Paul, R.M. Van Frank, W.L. Muth, J.W. Ross, D.C. Williams,Immunocytochemical demonstration of human proinsulin chimericpolypeptide within cytoplasmic inclusion bodies of Escherichia coli, Eur. J.Cell Biol. 31 (1983) 171–174.

[30] F. Baneyx, Recombinant protein expression in Escherichia coli, Curr. Opin.Biotechnol. 10 (1999) 411–421.

[31] D.B. Smith, K.S. Johnson, Single-step purification of polypeptides expressed inEscherichia coli as fusions with glutathione S-transferase, Gene 67 (1988) 31–40.

[32] M. Hammarstrom, N. Hellgren, S. van Den Berg, H. Berglund, T. Hard, Rapidscreening for improved solubility of small human proteins produced as fusionproteins in Escherichia coli, Protein Sci. 11 (2002) 313–321.

[33] P. Braun, Y. Hu, B. Shen, A. Halleck, M. Koundinya, E. Harlow, J. LaBaer,Proteome-scale purification of human proteins from bacteria, Proc. Natl. Acad.Sci. USA 99 (2002) 2654–2659.

[34] Y.P. Shih, W.M. Kung, J.C. Chen, C.H. Yeh, A.H. Wang, T.F. Wang, High-throughput screening of soluble recombinant proteins, Protein Sci. 11 (2002)1714–1719.

[35] M. Burg, J. Müthing, Characterization of cytosolic sialidase from Chinesehamster ovary cells: part I: cloning and expression of soluble sialidase inEscherichia coli, Carbohydr. Res. 330 (2001) 335–346.

[36] V. Thiemann, C. Donges, S.G. Prowe, R. Sterner, G. Antranikian, Characterisationof a thermoalkali-stable cyclodextrin glycosyltransferase from the anaerobicthermoalkaliphilic bacterium Anaerobranca gottschalkii, Arch. Microbiol. 182(2004) 226–235.

[37] B.H. Ha, K.J. Cho, Y.J. Choi, K.Y. Park, K.H. Kim, Characterization of argininedecarboxylase from Dianthus caryophyllus, Plant Physiol. Biochem. 42 (2004)307–311.

[38] K.N. Pandey, J. Kanungo, Expression of extracellular ligand-binding domain ofmurine guanylate cyclase/atrial natriuretic factor receptor cDNA in Escherichiacoli, Biochem. Biophys. Res. Commun. 190 (1993) 724–731.

[39] J.W. Kyle, P.J. Minogue, B.C. Thomas, D.A. Domowicz, V.M. Berthoud, D.A.Hanck, E.C. Beyer, An intact connexin N-terminus is required for function butnot gap junction formation, J. Cell Sci. 121 (2008) 2744–2750.