doi:10.1016/j.jmb.2011.07.013 J. Mol. Biol. (2011) 412, 72–79

Contents lists available at www.sciencedirect.com

Journal of Molecular Biologyj ourna l homepage: ht tp : / /ees .e lsev ie r.com. jmb

Two-Dimensional Kinetics of Inter-ConnexinInteractions from Single-Molecule Force Spectroscopy

Felix Rico1, Atsunori Oshima2, Peter Hinterdorfer3,Yoshinori Fujiyoshi2 and Simon Scheuring1⁎1Institut Curie, U1006 INSERM, 26 rue d'Ulm, 75005 Paris, France2Department of Biophysics, Faculty of Science, Kyoto University, Oiwake, Kitashirakawa, Sakyo-ku,Kyoto 606-8502, Japan3Institute for Biophysics, Johannes Kepler University of Linz, A-4040 Linz, Austria

Received 4 May 2011;received in revised form7 July 2011;accepted 11 July 2011Available online23 July 2011

Edited by C. R. Matthews

Keywords:gap junctions;connexins;adhesion;2D kinetics;atomic force microscopy

*Corresponding author. E-mail addsimon.scheuring@curie.fr.Abbreviations used: Cx, connexin

dimensional; AFM, atomic force micdynamic force spectroscopy; PEG, p

0022-2836/$ - see front matter © 2011 E

Gap junction channels are intercellular channels that form by docking theextracellular loops of connexin protein subunits. While the structure andfunction of gap junctions as intercellular channels have been characterizedusing different techniques, the physics of the inter-connexin interactionremain unknown. Moreover, as far as we know, the capacity of gap junctionchannels to work as adhesion complexes supporting pulling forces has notyet been quantitatively addressed. We report the first quantitativecharacterization of the kinetics and binding strength of the interaction ofa short peptide mimicking extracellular loop 2 of Cx26 with membrane-reconstituted Cx26, combining the imaging and force spectroscopycapabilities of atomic force microscopy. The fast dissociation rate inferreda dynamic bond, while the slow association rate reflected the reducedflexibility and small size of extracellular loops. Our results propose the gapjunction channel as an adhesion complex that associates slowly anddissociates fast at low force but is able to support important pulling forces inits native, hexameric form.

© 2011 Elsevier Ltd. All rights reserved.

Introduction

Gap junction intercellular communication exists invirtually all multicellular organisms. Gap junctionchannels are formed by connexin (Cx) proteins,which form hexameric barrels termed connexons or“hemichannels.” Two connexons from opposingcells dock coaxially to form a gap junction channelthrough the interaction of the extracellular loops ofeach Cx.1,2 Cx proteins exist in different isoforms,such as Cx26 and Cx46, important Cxs for theproper function of the inner ear and the eye lens,

ress:

; 2D, two-roscopy; DFS,olyethylene glycol.

lsevier Ltd. All rights reserve

respectively.3 Gap junctions allow the exchange ofions and molecules of size up to 1 kDa,4 and theirinitial characterization has been mainly restricted tothis channeling function. However, recent worksdemonstrated that gap junctions also function asadhesion molecules in essential processes.5,6 Despitethe adhesive function of gap junction channels, littleis known about the docking mechanism of connexonpairs. A great step forward in the understanding ofgap junction structure and function has beenachieved by the recent 3.5-Å structure determinationusing X-ray crystallography that described theinteraction surfaces at the atomic level. Opposingconnexons dock to form intercellular channels byinteractions mediated by the two extracellular loopsof each Cx, which have homologous sequencesamong β-type human Cxs (Fig. 1a).1 Among theresidues that mediate inter-connexon interaction, anexposed stretch of amino acids NTVD (Fig. 1b)

d.

NT

VD

(a)

(b)

(d)

(c)(e)

(f)PEG

NTVD

membrane-2

membrane-1

mica support/piezoelectric stage

lase

r

50 nm

12

8

4

0

Hei

ght (

nm)

lipid

lipid

Cx26-extracellular

Cx26-extracellular

Cx26-cytoplasmic

AFM

tip

photo

dio

de

Fig. 1. Cx26 structure and experimental setup. (a) Structure of an individual Cx26 in ribbon representation. (b) Closeview of extracellular loop 2 (E2 loop) showing the side chains of residues NTVD that interact with the E2 loop of theopposing Cx. (c) Close view of two E2 loops from opposing Cx26 with highlighted side chains of interacting residues. (d)Schematic representation of the experimental setup showing the mica on the piezoelectric stage support where 2D crystalsof Cx26 were deposited and the AFM cantilever tip with NTVD peptides attached through a PEG linker. The deflection ofthe cantilever was detected in a segmented photodiode by reflecting a laser beam off the backside of the cantilever. (e)Topographical image of Cx26 2D crystals showing two layers where individual connexons are recognized. The false colorscale represents 16 nm. (f) Topographical cross-section along the broken line in (e). The scale bar in (e) is also valid.

73Kinetics of Inter-Connexin Interactions

within extracellular loop 2 (E2 loop) has beenidentified to be responsible for the selective speci-ficity of Cxs (Fig. 1c).7,8 In earlier attempts toelucidate the molecular determinants of gap junc-tion formation and function, peptides mimickingshort sequences of Cxs have been shown tospecifically inhibit gap junction formation.9–11

While the kinetics and binding strength of manyadhesion molecules have been determined usingvarious techniques, the binding properties of thegap junction interaction are still unknown. Theformation of gap junctions is a membrane-mediatedphenomenon and requires direct physical contactbetween the two-dimensional (2D) surfaces ofneighboring cells. Therefore, a reliable determina-tion of the binding affinity of gap junctions wouldrequire in situ determination of receptor and liganddensities together with a technique capable ofmeasuring 2D binding kinetics.12

Results and Discussion

To determine the 2D kinetics and binding strengthof the inter-Cx interaction, we have combined theimaging and force spectroscopy capabilities ofatomic force microscopy (AFM).13 We decoratedAFM tips with the gap junction mimetic peptide tocarry out force spectroscopy measurements on 2Dcrystals of hCx26M34A (human Cx26 with single-site mutation M34A). While intermittent contactimaging allowed the in situ determination of theprotein density, essential to reliably determine 2Dkinetics, dynamic force spectroscopy (DFS) andbinding probability plots were used to determinethe 2D kinetics and binding strength of the inter-Cxinteraction in the presence and absence of Ca2+. Ourresults provide unique information about themechanism of binding, maintenance and dissocia-tion of gap junction channels.

74 Kinetics of Inter-Connexin Interactions

Essential residues in extracellular loop 2 (E2 loop)interact with the opposing Cx monomer in the gapjunction channel structure of Cx26 (Fig. 1a). Wetarget in our experimental rationale a specific NTVDsite in E2 (Fig. 1b) that pairs with the identicalstretch in the opposing Cx (Fig. 1c). The mimeticpeptide LGGGGGNTVD was bound through aflexible polyethylene glycol (PEG) linker to theAFM tip to allow orientation freedom and distanceto the attachment point (Fig. 1d).14 We used AFM inPeakForce imaging mode15 to characterize mem-branes with densely reconstituted hCx26M34A. Weselected, by AFM imaging, reconstituted mem-branes that presented only two layers and exposedconnexons reconstituted in both orientations (Fig.1e). In topographical images of Cx26 membranes,individual connexons were resolved, which wasremarkable considering that the scanning processwas performed with a functionalized tip (Fig. 1e).The small nominal radius of the used AFM tips(2 nm) and the density of peptides allowed us todiscern individual connexons, although the resolu-tion was not sufficient to resolve the central cavity.The brighter circles in two different superposedmembrane layers in a densely packed organizationrepresent the extracellular face exposing the E2loops. The average density of connexons was(3.7±0.6)×10−3 connexons/nm2, leading to 0.02 E2loops/nm2. It is important to mention that thedensity of connexons we report here was only afactor of 3 lower than that found on nativemembranes. For example, in native eye lensmembranes,16 connexons were closely packed withan average inter-connexon distance of ∼10 nm,leading to a density of 0.06 E2 loops/nm2. More-over, freeze–fracture studies in liver membranesshowed gap junction plaques with connexons in a2D lattice, similar to that observed here.17 Cross-section of AFM topographs (Fig. 1f) revealed thatthe height of each layer including the proteins was∼8 nm, with the extracellular portion protruding∼2.3 nm above the lipid surface, in good agreementwith the dimensions from electron microscopy andX-ray analyses.1,2 PeakForce AFM imaging allowedus to place the functionalized tip onto a desiredmembrane region where connexons exposed theirextracellular face without damaging the tip modifi-cation because, in PeakForce imaging, loading forcescan be minimized to ∼50 pN and friction forces areeliminated.15

On the extracellular surface, we acquired largenumbers (nN20,000) of force–distance curves atdifferent velocities and under different conditions.Rupture events from retraction curves were detectedin ∼10% of the curves. The assignment of specificevents was further subjected to the criterion offinding nonlinear force-extension profiles beforerupture, characteristic of PEG stretching, welldescribed by the extended freely jointed chain

model (Fig. 2a).18 The rupture force was determinedas the difference between the peak adhesion forceand the baseline after detachment (Fig. 2a, blackarrow). The most probable rupture forces at eachloading rate were determined from histograms(Fig. 2b) and were used to generate the dynamicforce spectra (rupture force versus loading rate) ofthe interaction under Ca2+-free and 1-mM Ca2+

conditions. Rupture forces showed a linear depen-dence with the logarithm of the loading rate,suggesting a single dissociation barrier. The addi-tion of 1 mM Ca2+ in the measurement buffersignificantly increased the slope of the resultingdynamic force spectrum of the interaction(p=0.029, Fig. 2). We fitted a model describingthe forced dissociation of molecular bonds to thedynamic force spectra [Eq. (3)]19 to determine thedissociation rate (koff), potential width (xβ) andpotential depth (ΔG) of the interaction under eachcondition (Table 1). In addition to the ∼3.5-foldincrease in the dissociation rate under 1-mM Ca2+

conditions, the decrease in the potential width ofthe interaction, which indicated tightening of thebond, is noteworthy.The determination of the 2D binding affinity (Ka)

requires knowledge of the ligand and receptordensities and measurement of the adhesion proba-bility as a function of the contact time.12 Unlike themore common three-dimensional binding affinity,which has units of volume, the 2D binding affinityhas units of area, more appropriate for membrane-associated processes.12 As expected, the bindingprobability (percentage of curves with ruptureevents) increased with increased contact timebetween the two surfaces, tending to a plateauvalue (Fig. 2d). The application of Ca2+ signifi-cantly decreased the plateau value of the bindingprobability to about 0.5, compared to 0.8 in theabsence of Ca2+ (p=0.046). The binding affinity(Ka) was determined by modeling the bindingprobability as a forward reaction of second orderand a reverse reaction of first order [Eq. (S4)](Table 1).12 Ka was ∼4-fold higher under Ca2+-freeconditions and was used to derive the associationrates using kon=Ka×koff. Given the lower dissoci-ation rate under Ca2+-free conditions, the derivedassociation rates were very similar for the twoconditions (Table 1).Our results suggest that Ca2+ has influence during

dissociation but not during association, beingslightly easier to dissociate at zero force whilebeing more resistant to applied force load. It isknown that Ca2+ affects gap junction formation andinduces hemichannel closure.8,9,20 A recent electroncrystallographic study revealed that a cytoplasmicplug in the vestibule may regulate channel gating.2

Alternatively, an AFM study reported conforma-tional changes in the extracellular region in thepresence of Ca2+.21 The contraction of the pore

Bin

ding

pro

babi

lity

fr

Rup

ture

forc

e (p

N)

102 103 104 105

Ca2+

free 1 mM Ca2+

Ca2+

free 1 mM Ca2+

Pro

babi

lity

dens

ity (

a.u.

)

Force (pN)

Rel

ativ

e ad

h. fr

eq.

control sol. NTVD lipid

For

ce

Tip displacement

5 nm

50 pN

1.0

0.8

0.6

0.4

0.2

0.0

550 pN/s

3900 pN/s

94200 pN/s

3002001000

250

200

150

100

50

0

106 1.00.80.60.40.20.0Contact time (s)Loading rate (pN/s)

1.0

0.8

0.6

0.4

0.2

0.0

(b)(a)

(d)(c)

Fig. 2. 2D kinetics. (a) Representative example of a retraction force curve obtained from force measurements underCa2+-free conditions presenting a single rupture event. The characteristic extended, freely jointed chain model (red line)was a signature of the specificity of the interaction. Rupture forces (fr, arrow) and instantaneous loading rates wereextracted from retraction curves. The specificity of the interaction (inset) was further confirmed from the drop in therelative adhesion frequency after adding soluble NTVD to the measurement buffer (black bar) and carrying outexperiments on the lipid surface (gray bar). Specificity measurements were carried out using the same tip and sampleunder the various conditions on three independent measurements. (b) Rupture force histograms at three different loadingrates (an average of 140 rupture events per histogram was used). (c) The most probable rupture force (±standarddeviation) from each histogram was represented versus the loading rate to generate the dynamic force spectrum of theinteraction for Ca2+-free (open symbols) and 1-mM Ca2+ (filled symbols) conditions. Lines show the best fits of Eq. (3)used to extract the parameters of the interaction. (d) Binding probability (mean±standard error of the mean, N=3) versuscontact time for Ca2+-free (open symbols) and 1-mM Ca2+ (filled symbols) conditions. Lines represent the best fits of Eq.(S4) used to extract the binding affinities.

Table 1. NTVD/E2 loop interaction parameters

Condition

Dissociation Association

koff(1/s)

xβ(nm)

ΔG(kBT)

Ka0

(nm2)kon

(nm2/s)

Ca2+ free 0.8±0.7 0.47±0.13 12.7±4.6 126±6 101±171 mM Ca2+ 2.8±0.8 0.28±0.04 9.0±1.5 30±2 85±5

koff, dissociation rate; xβ, potential width;ΔG, potential depth; Ka0,

binding affinity; and kon, association rate.

75Kinetics of Inter-Connexin Interactions

radius observed by AFM of Cx26 crystals in thepresence of Ca2+ could be due to stiffening of theextracellular loops. Thus, the change we observed inthe dissociation pathway could well be explainedby a modification of the flexibility of the E2 loop,which may vary the energy landscape of the inter-action. In addition, it has been suggested thatextracellular Ca2+ at micromolar concentrationsmay help to stabilize hydrophobic interactionsbetween hemichannels.8 The force dependence ofthe dissociation rate is shown in Fig. 3a and predicts

Dis

soci

atio

n ra

te (

1/s)

Force (pN)

N=1

Ca2+ free

1mM Ca2+

N=6

Ca2+ free

1mM Ca2+

2 kBT

0.2 nm

Fre

e en

ergy

Reaction coordinate

100

101

102

103

104

16012080400

(b)(a)

Fig. 3. Dissociation rate under force and energy landscape. (a) Predicted dissociation rates under applied force forCa2+-free (black lines) and 1-mM Ca2+ (red lines) conditions for a single NTVD/E2 loop bond (continuous lines) andhexameric (dotted lines) bonds loaded in parallel. (b) Sketch of the energy landscape of the single NTVD/E2 loopinteraction from the parameters provided in Table 1.

76 Kinetics of Inter-Connexin Interactions

that, at low forces (below 30 pN), the dissociation isslower under Ca2+-free conditions than in thepresence of 1 mM Ca2+. Due to the smaller potentialwidth of the interaction, beyond this force level, thecomplex dissociates slower in the presence of Ca2+.This behavior suggests that Ca2+ may stabilize thecomplex when subjected to high pulling forces; it isuncertain, though, if single inter-Cx bonds aresubjected to such forces in vivo. The sketch of thefree-energy landscape summarizes the NTVD/E2loop interaction (Fig. 3b). While at low forces, thedissociation barrier of the complex is lower in thepresence of Ca2+ than under Ca2+-free conditions andthus dissociates faster, the steeper energy barrierinduces amore stable bond at high forces. By contrast,a similar height of the activation barrier of associationreflects the similar values of the association rates.The 2D kinetic parameters of the NTVD/E2 loop

interaction are comparable to those reported for theclose and intermediate states of integrin αLβ2,implying a low-affinity complex.22 In addition, thedissociation rates and potential widths are similar tothe values found in selectins, suggesting a rather fastdissociation.23 We can conclude that the NTVD/E2loop interaction is governed by a relatively slowbinding rate and a fast dissociation rate. A slowassociation rate may reflect the limited lateralmobility not only of hemichannels within themembrane in the densely packed configuration ofour reconstituted membranes but also of native gapjunction plaques, as found in liver17 and in eyelens.16 Furthermore, the extracellular loops of Cx26protrude only ∼2 nm and are cross-linked by threedisulfide bridges; it is thus expected to be a rigidstructure with restricted flexibility also slowing

down association.1,23 In contrast, the fast dissocia-tion rates are in accordance with recent workshowing that Cxs are able to mediate cellmigration.6 Thus, our results support the hypothesisof gap junctions as stable complexes, although withinter-Cx bonds in dynamic equilibrium.We have used a short peptide to mimic the

homotypic interaction of the extracellular E2 loopof Cx26. Although it represents a single E2–E2interaction, our results represent the first quantita-tive assessment of the gap junction extracellularinteraction. Besides the likely additional contribu-tion of extracellular loop 1, the kinetics of theinteraction are expected to be different in the actualgap junction channel configuration, where sixcomplexes are formed simultaneously. On the onehand, given the highly oriented and orderedorganization of the extracellular loops, we expectcooperative binding. The cooperation of adhesionmolecules by clustering has been suggested to be animportant mechanism of affinity modulation inleukocytes.12 Thus, upon binding of one Cx pair, itis expected that the unbound monomers of the samehemichannel would have already a proper orienta-tion, increasing the binding propensity. On the otherhand, recent works have shown this behavior ofmultimeric complexes under applied force.24,25

Assuming the fully formed gap junction channel assix identical receptor/ligand complexes subjected toan external force and using the parameters obtainedfor the NTVD/E2 loop interaction, we have com-puted [Eq. (2)] the dissociation rate of the hexamericcomplex under constant pulling forces (Fig. 3a). Theeffect of force on the hexameric complex is much lesspronounced, and the level of force at which Ca2+

77Kinetics of Inter-Connexin Interactions

leads to a slower rate rises to ∼130 pN. Thus,although the dissociation rate at zero force of theconnexon pair is relatively fast, the hexamericcomplex is able to support important forces beforedissociation, comparable to that supported byintegrins in the high-affinity state.26,27 Confirmationof this behavior on the actual connexon–connexoninteraction by a similar approach will establish gapjunction channels as force-bearing complexes thatmay contribute importantly to the structural inte-grity of cell–cell contacts.

Materials and Methods

Protein purification and 2D crystallization ofhCx26M34A hemichannels

2D crystals of hCx26M34A were prepared as described2,28

with minor modifications. Briefly, purified hCx26M34Ahemichannels were mixed with decyl-maltoside-solubilizeddioleoylphosphatidylcholine (Avanti Polar Lipids, Alabas-ter, AL) at a lipid-to-protein ratio of 1. The mixture wasdialyzed against 10 mM 4-morpholineethanesulfonic acid(pH 5.8), 100 mM NaCl, 50 mM MgCl2, 5 mM CaCl2,2 mM DTT, 100 μM carbenoxolone (Sigma, St. Louis,MO), 0.005% NaN3 and 1% glycerol. The concentration ofNaCl was increased to 300 mM 72 h later. Dialysis wasperformed at 20 °C for the first 24 h, 37 °C for the next96 h, and 20 °C for the final 24 h. The dialysis solutionwas exchanged once after temperature was increased to37 °C.

Sample preparation for AFM measurements

Prior to physisorption on mica for AFM analysis, triple-layered Cx26 2D crystals2,28 were incubated for 15 min in adissociation buffer that consisted of 10 mM Tris–HCl atpH 8.0, 5 mM ethylenediaminetetraacetic acid and 5 mMethylene glycol bis(β-aminoethyl ether) N,N′-tetraaceticacid to favor layer dissociation and exposure of theextracellular faces of Cx26. 2D crystals of hCx26M34Awere then immobilized on freshly cleaved mica byphysisorption during incubation for 1 h in adsorptionbuffer (10 mM Tris–HCl at pH 7.4, 150 mM KCl and25mMNiCl2) at room temperature. Before measurements,samples were rinsed with measurement buffer (10 mMTris–HCl at pH 7.4 and 150 mM KCl) with or without1 mM CaCl2.

Immobilization of mimetic peptides on AFM tips

Mimetic peptides (99.3% purity as determined fromHPLC analysis) consisted of a lysine followed by fiveglycines before the NTVD motif of the E2 loop (aminoacids 175–179 in the extracellular loop 2 of human Cx26)and were obtained from PolyPeptide Group (France).Si3N4 cantilevers with silicon tips (MSNL; Veeco,

Santa Barbara, CA) were washed in chloroform, driedunder nitrogen stream and incubated overnight inethanolamine solution (0.5 g/ml in dimethyl sulfoxide).

Cantilevers were then washed in dimethyl sulfoxide andethanol and dried under nitrogen stream. Ethylamine-coated cantilevers were immersed in 6.6 mg/ml aldhe-hyde-PEG-NHS (obtained from Prof. Hermann Gruber,Johannes Kepler University, Linz, Austria) solution inchloroform with 30 μl of triethylamine [final concentra-tion of 0.5% (vol/vol)] and incubated for 2 h at roomtemperature. Then, cantilevers were washed with chlo-roform and dried. Cantilevers were then immersed in adrop of peptide solution (2.3 mg/ml) with 2 μl of 1 Msodium cyanoborohydride and incubated for 1 h atroom temperature. To passivate unreacted aldehydegroups, we added 5 μl of 1 M ethanolamine and let itincubate for 10 min at room temperature. Cantileverswere finally rinsed with Tris–HCl buffer and stored at4 °C until used.29

AFM measurements

AFM measurements were performed in buffer solutionat room temperature and ambient pressure on a commer-cial Nanoscope-V AFM (Veeco) equipped with Nanoscope8 control software, vertical engagement, 160-μm scanner(J-scanner) and Si3N4 cantilevers with silicon tips and anominal spring constant of 100 pN/nm and a 2-nmnominal radius.The spring constant of the cantilevers was first

calibrated using the thermal fluctuation method.30 Afterlocalization of the 2D crystals by imaging the samplesurface in PeakForce mode, the tip was positioned on thereconstituted membrane, and force measurements wereacquired.Force spectroscopy measurements consisted of force

curves obtained by bringing the mimetic-peptide-coatedtip into contact with the Cx26 2D crystal with a contactduration of 10 ms and a compression force of ∼200 pNand then retracting the sample at various speeds, from0.03 to 10 μm/s. Adhesion events were only observed in∼10% of the retraction force curves, which ensured thatN95% probability that the adhesion event was mediatedby a single ligand/receptor bond.31 The unbinding forces(fr) of individual ligand/receptor complexes were derivedfrom the jump in force following the separation of thecantilever from the sample. Instantaneous loading rates(rf) were determined from the slope before rupture.Rupture forces were pooled by loading rate, and histo-grams were generated. The most probable force was thenrepresented as a function of loading rate to generate thedynamic force spectra.Binding probability measurements consisted of acquir-

ing force curves at constant approaching and retractionspeeds of 1 μm/s varying the contact duration from 6 msto 1 s. This upper limit was used to minimize the numberof unspecific binding, which increases at long contacttimes. The binding probability was computed by dividingthe number of curves presenting rupture events by thetotal number of curves.

Data analysis

Our unbinding data were analyzed in terms of anextension of the Bell–Evans model that considers theshape of the energy landscape.19,32 When a force (F) is

78 Kinetics of Inter-Connexin Interactions

applied to the bond, the energy landscape is distorted,enhancing the dissociation rate

k Fð Þ = koff 1−mFxhDG

� �1= m−1

expDGkBT

1 − 1−mFxhDG

� �1= m" #( )

ð1Þwhere koff, xβ and ΔG are the intrinsic dissociation rate,width and height of the interaction potential, respectively;kB is the Boltzmann constant; and T is the absolutetemperature. The dissociation rate of six identical recep-tor/ligand complexes subjected to an external force (F) hasbeen derived to be

kN =6 Fð Þ =XNn=1

1nk F

n

� �" #−1

ð2Þ

where k(F/n) is the dissociation constant of a single bond[Eq. (1)] under a force shared by n bonds.33

Assuming a harmonic potential (ν=1/2), the mostprobable rupture force can be described as a function ofthe loading rate (rf) by

19

FiDGmxh

1 −kBTDG

1nkoffkBTe

DGkBT

+ 0:577

xhrf

!m" #ð3Þ

To estimate the interaction parameters, we fitted Eq. (3)to the force spectra.Binding probability plots were fitted with a probabilis-

tic kinetic model described by

Pa = 1 − exp 1 − mrmlAcK0a 1 − exp −kofftð Þ½ �� � ð4Þ

where the surface densities of receptors (mr, Cx26) andligands (ml, peptides) and the area of contact between thetip and the sample (Ac) were determined from theexperimental conditions and the intrinsic dissociationrate koff was determined from fitting Eq. (3) to the DFS.The binding affinity (Ka

0) was the only fitting parameter.The density of receptors (E2 loops) was determined fromthe AFM topographs. The effective density of ligands (ml)was determined by considering a single molecule at theend of the tip covering an area defined by a spherical capof effective radius reff (PEG linker plus peptide, ∼6 nm).The area of contact (Ac) was then assumed to be that samearea, which led to mlAc=1 in Eq. (4).13,29

Statistics

Unless otherwise stated, data were represented asmean±standard deviation. The combined effects ofCa2+ and loading rate or contact time in the DFS or bind-ing probability plots were tested by two-way analysisof variance.

Acknowledgements

This study was supported by the Agence Natio-nale de la Recherche and the “City of Paris” (to S.S.)

and a European Community Marie Curie Intra-European Fellowship for Career Development (toF.R.). The authors thank Dr. William Silverman forinsightful discussions and Dr. Lilia A. Chtcheglovafor technical assistance.

References

1. Maeda, S., Nakagawa, S., Suga, M., Yamashita, E.,Oshima, A., Fujiyoshi, Y. & Tsukihara, T. (2009).Structure of the connexin 26 gap junction channel at3.5 Å resolution. Nature, 458, 597–602.

2. Oshima, A., Tani, K., Hiroaki, Y., Fujiyoshi, Y. &Sosinsky, G. E. (2007). Three-dimensional structure ofa human connexin26 gap junction channel reveals aplug in the vestibule. Proc. Natl Acad. Sci. USA, 104,10034–10039.

3. Beyer, E. C. & Berthoud, V. M. (2009). Chapter 1: TheFamily of Connexin Genes. In Connexins: A Guide(Harris, A. L. & Darren, eds), pp. 3–26, Springer, NewYork, NY.

4. Kumar, N. M. & Gilula, N. B. (1996). The gap junctioncommunication channel. Cell, 84, 381–388.

5. Cotrina, M. L., Lin, J. H. & Nedergaard, M. (2008).Adhesive properties of connexin hemichannels. Glia,56, 1791–1798.

6. Elias, L. A., Wang, D. D. & Kriegstein, A. R. (2007).Gap junction adhesion is necessary for radial migra-tion in the neocortex. Nature, 448, 901–907.

7. Bruzzone, R., White, T. W. & Paul, D. L. (1994).Expression of chimeric connexins reveals new prop-erties of the formation and gating behavior of gapjunction channels. J. Cell Sci. 107, 955–967.

8. Harris, A. L. (2001). Emerging issues of connexinchannels: biophysics fills the gap. Q. Rev. Biophys. 34,325–472.

9. Dahl, G., Werner, R., Levine, E. & Rabadandiehl, C.(1992). Mutational analysis of gap junction formation.Biophys. J. 62, 172–182.

10. Dahl, G. (2007). Gap junction–mimetic peptides dowork, but in unexpected ways. Cell Commun. Adhes.14, 259–264.

11. Evans, W. H. & Leybaert, L. (2007). Mimetic peptidesas blockers of connexin channel-facilitated intercellu-lar communication. Cell Commun. Adhes. 14, 265–273.

12. Huang, J., Zarnitsyna, V. I., Liu, B., Edwards, L. J.,Jiang, N., Evavold, B. D. & Zhu, C. (2010). The kineticsof two-dimensional TCR and pMHC interactionsdetermine T-cell responsiveness. Nature, 464, 932–936.

13. Hinterdorfer, P., Baumgartner, W., Gruber, H. J.,Schilcher, K. & Schindler, H. (1996). Detection andlocalization of individual antibody–antigen recogni-tion events by atomic force microscopy. Proc. NatlAcad. Sci. USA, 93, 3477–3481.

14. Ebner, A., Wildling, L., Kamruzzahan, A. S., Rankl, C.,Wruss, J., Hahn, C. D. et al. (2007). A new, simplemethod for linking of antibodies to atomic forcemicroscopy tips. Bioconjugate Chem. 18, 1176–1184.

15. Su, C. (2010). Mapping quantitative mechanicalproperties at molecular scale using peak force tappingAFM. Microsc. Microanal. 16, 364–365.

16. Buzhynskyy, N., Hite, R. K., Walz, T. & Scheuring, S.(2007). The supramolecular architecture of junctional

79Kinetics of Inter-Connexin Interactions

microdomains in native lens membranes. EMBO Rep.8, 51–55.

17. Goodenough, D. A. & Revel, J. P. (1970). A finestructural analysis of intercellular junctions in themouse liver. J. Cell Biol. 45, 272–290.

18. Baumgartner, W., Hinterdorfer, P., Ness, W., Raab, A.,Vestweber, D., Schindler, H. & Drenckhahn, D. (2000).Cadherin interaction probed by atomic force micros-copy. Proc. Natl Acad. Sci. USA, 97, 4005–4010.

19. Dudko, O. K., Hummer, G. & Szabo, A. (2006).Intrinsic rates and activation free energies fromsingle-molecule pulling experiments. Phys. Rev. Lett.96, 108101.

20. Beahm, D. L. & Hall, J. E. (2004). Opening hemi-channels in nonjunctional membrane stimulates gapjunction formation. Biophys. J. 86, 781–796.

21. Muller, D. J., Hand, G. M., Engel, A. & Sosinsky, G. E.(2002). Conformational changes in surface structuresof isolated connexin 26 gap junctions. EMBO J. 21,3598–3607.

22. Zhang, F., Marcus, W. D., Goyal, N. H., Selvaraj, P.,Springer, T. A. & Zhu, C. (2005). Two-dimensionalkinetics regulation of Œ±LŒ≤2-ICAM-1 interactionby conformational changes of the Œ±L-inserteddomain. J. Biol. Chem. 280, 42207–42218.

23. Huang, J., Chen, J., Chesla, S. E., Yago, T., Mehta, P.,McEver, R. P. et al. (2004). Quantifying the effects ofmolecular orientation and length on two-dimensionalreceptor–ligand binding kinetics. J. Biol. Chem. 279,44915–44923.

24. Sulchek, T. A., Friddle, R. W., Langry, K., Lau, E. Y.,Albrecht, H., Ratto, T. V. et al. (2005). Dynamic forcespectroscopy of parallel individual Mucin1–antibodybonds. Proc. Natl Acad. Sci. USA, 102, 16638–16643.

25. Rankl, C., Kienberger, F., Wildling, L., Wruss, J.,Gruber, H. J., Blaas, D. & Hinterdorfer, P. (2008).Multiple receptors involved in human rhinovirusattachment to live cells. Proc. Natl Acad. Sci. USA, 105,17778–17783.

26. Zhang, X. H., Wojcikiewicz, E. & Moy, V. T. (2002).Force spectroscopy of the leukocyte function-associ-ated antigen-1/intercellular adhesion molecule-1 in-teraction. Biophys. J. 83, 2270–2279.

27. Zhang, X. H., Craig, S. E., Kirby, H., Humphries, M. J.& Moy, V. T. (2004). Molecular basis for the dynamicstrength of the integrin alpha(4)beta(1)/VCAM-1interaction. Biophys. J. 87, 3470–3478.

28. Oshima, A., Doi, T., Mitsuoka, K., Maeda, S. &Fujiyoshi, Y. (2003). Roles of Met-34, Cys-64, andArg-75 in the assembly of human connexin 26. J. Biol.Chem. 278, 1807–1816.

29. Riener, C. K., Stroh, C. M., Ebner, A., Klampfl, C., Gall,A. A., Romanin, C. et al. (2003). Simple test system forsingle molecule recognition force microscopy. Anal.Chim. Acta, 479, 59–75.

30. Hutter, J. L. & Bechhoefer, J. (1993). Calibration ofatomic-force microscope tips. Rev. Sci. Instrum. 64,1868–1873.

31. Evans, E., Kinoshita, K., Simon, S. & Leung, A.(2010). Long-lived, high-strength states of ICAM-1bonds to beta2 integrin, I: lifetimes of bonds torecombinant alphaLbeta2 under force. Biophys. J. 98,1458–1466.

32. Bell, G. I. (1978). Models for specific adhesion of cellsto cells. Science, 200, 618–627.

33. Williams, P. M. (2003). Analytical descriptions ofdynamic force spectroscopy: behaviour of multipleconnections. Anal. Chim. Acta, 479, 107–115.