Molecular interaction and functional regulation of connexin50 gap junctions by calmodulin

Abstract

Cx50 (connexin50), a member of the α-family of gap junction proteins expressed in the lens of the eye, has been shown to be essential for normal lens development. In the present study, we identified a CaMBD [CaM (calmodulin)-binding domain] (residues 141–166) in the intracellular loop of Cx50. Elevations in intracellular Ca²⁺ concentration effected a 95 % decline in g_j (junctional conductance) of Cx50 in N2a cells that is likely to be mediated by CaM, because inclusion of the CaM inhibitor calmidazolium prevented this Ca²⁺-dependent decrease in g_j. The direct involvement of the Cx50 CaMBD in this Ca²⁺/CaM-dependent regulation was demonstrated further by the inclusion of a synthetic peptide encompassing the CaMBD in both whole-cell patch pipettes, which effectively prevented the intracellular Ca²⁺-dependent decline in g_j. Biophysical studies using NMR and fluorescence spectroscopy reveal further that the peptide stoichiometrically binds to Ca²⁺/CaM with an affinity of ~ 5 nM. The binding of the peptide expanded the Ca²⁺-sensing range of CaM by increasing the Ca²⁺ affinity of the C-lobe of CaM, while decreasing the Ca²⁺ affinity of the N-lobe of CaM. Overall, these results demonstrate that the binding of Ca²⁺/CaM to the intracellular loop of Cx50 is critical for mediating the Ca²⁺-dependent inhibition of Cx50 gap junctions in the lens of the eye.

INTRODUCTION

The circulation of current fluid flow through the closely packed cells of the lens of the eye is essential for lens homoeostasis, and is highly dependent on functional cell–cell coupling. This is accomplished by gap junctions, relatively non-selective channels that are permeable to low-molecular-mass (<1 kDa) molecules that directly link the cytoplasms of adjacent cells [1–3]. The gap junction channel formed between two adjacent cells comprises two connexon hemichannels, one from each cell, joined in mirror symmetry. Each hemichannel is composed of six transmembrane Cx (connexin) proteins embedded in the plasma membrane of the cell. The mammalian lens comprises three different Cx proteins, namely Cx43, Cx46 and Cx50, that, like all other Cxs, have four transmembrane segments, a short N-terminal cytoplasmic region, one intracellular loop and two extracellular loops with a set of highly conserved cysteine residues; the major difference between all Cxs is in the sequences of their cytoplasmic loops and C-terminal tails [3].

Inhibition of cell–cell communication in lens by elevated [Ca²⁺]_i (intracellular Ca²⁺ concentration) was first demonstrated as an increased internal electrical resistance [4] that was prevented by pre-incubation with CaM (calmodulin) antagonists. Elevated [Ca²⁺]_i also inhibits cell–cell communication in both bovine [5] and sheep [6] lens primary cell cultures. We have demonstrated that lens cell–cell communication was half-maximally inhibited at ~ 300 nM [Ca²⁺]_i [6] and was prevented by pre-incubation of lens cultures with CaM antagonists [7]. The rapid onset of this inhibition (within seconds) suggests that this is mediated by a direct interaction of CaM with one or more of the lens Cxs, rather than being mediated via the action of a CaM-dependent protein kinase. Indeed, Peracchia et al. [8] have demonstrated that CaM gates Cx32-containing gap junctions directly, and Török et al. [9] have identified two distinct CaM-binding amino acid sequences in Cx32, with the N- and C-lobes of CaM showing separate functions, suggesting trans-domain or trans-subunit bridging by CaM as a possible mechanism of gap junction gating [10]; notably, this CaM-binding motif in Cx32 is absent from the three lens Cxs.

To better understand the Ca²⁺-dependent inhibition of lens gap junctions, we identified high-affinity CaM-binding sites in the lens Cxs Cx43 and Cx44 (the sheep homologue of human Cx46). Specifically, these domains encompass residues 138–157 in Cx43 [11] and residues 129–148 in Cx44 (corresponding to residues 138–157 in Cx46) [12]. These sequences have conserved hydrophobic residues at positions 1, 5, 10 and 14 that matched well with two major CaM-binding classes, 1–10 and 1–14, which are similar to the CaM-binding motifs found in calcineurin, nitric oxide synthase, adenylate cyclase and skeletal myosin light chain kinase [13,14]. These CaMBDs (CaM-binding domains) are located at the C-terminal end of the loop region between the predicted Cx TM (transmembrane region) 2 and TM3 [2–3,15–18] and are highly conserved in all three lens Cxs.

The present study was conducted to determine whether Cx50, like Cx43 and Cx44, was also regulated by [Ca²⁺]_i in the micromolar range, and, if so, whether this Ca²⁺ regulation was CaM-mediated. A bioinformatic analysis [14] of Cx50 resulted in the identification of a candidate CaMBD with high predictive score that was located in the same C-terminal end of the loop region between the predicted Cx TM2 and TM3 of Cx50. This region is highly conserved in all three lens Cxs. A peptide model, which proved to be successful in our studies with Cx43 and Cx44 [11,12], was then adopted to demonstrate the physical contacts between the predicted Cx50 CaMBD and CaM. Studies using high-resolution NMR and fluorescence spectroscopy clearly established a high-affinity Ca²⁺-dependent interaction between a domain in the intracellular loop of Cx50 and CaM. Our results provide important new insights into the molecular mechanism by which altered regulation of intracellular Ca²⁺ concentration in the lens affects the regulation of lens Cx50 gap junctions.

EXPERIMENTAL

Prediction of transmembrane topology and a CaMBD

On the basis of a hidden Markov model, the topology and orientation of the transmembrane helices of the mouse Cx50 were predicted using four different programs, including TMHMM, MEMSAT, SOSUI and HMMTOP [16,19–21]. ClustalW2 was used to align multiple mammalian gap junction polypeptide sequences in the α-family, including Cx43 (GenBank^® accession number NP_000156; Gja1 human), Cx44 (GenBank^® accession number AAD56220; Gja3 sheep), Cx46 (GenBank^® accession number NP_068773; Gja3 human) and Cx50 (GenBank^® accession number AAF32309; Gja8 human). The candidate CaMBD in Cx50 was predicted using the CaM Target Database developed on the basis of over 100 known CaM target sequences [14].

Peptides and proteins

The Cx50 peptide, Cx50p^141–166 (S¹⁴¹SKGTKKFRLEGTLLRTYVCHIIFKT¹⁶⁶), was synthesized by EZ Biolab and purified by preparative reversed-phase HPLC with a purity of >87 %. A scrambled peptide, SCx50p (FKLYKCISFGGTEITTRSHVLTKKRL), with the same composition of amino acids which shows no predicted CaM-binding capacity, was synthesized with a purity of >90 % and served as a negative control. The molecular masses of these two peptides were determined by MALDI (matrix-assisted laser-desorption/ionization)–TOF (time-of-flight)-MS in reflectron mode. To mimic the protein environment and remove extra charges, peptides were acylated at their N-termini and aminated at their C-termini.

Both ¹⁴N and ¹⁵N isotopically labelled recombinant rat CaMs were expressed and purified as described previously [11]. D-CaM (dansylated CaM) was prepared as described previously [22]. The concentration of CaM was determined using the ε₂₇₆ of 3030 M⁻¹ · cm⁻¹ [23]. The amount of dye bound was determined using an ε₃₃₅ of 3980 M⁻¹· cm⁻¹ for D-CaM. The concentration of modified CaM was determined using the BCA (bicinchoninic acid) assay (Pierce) with BSA or unmodified CaM as the standard. The modification of CaM was confirmed further by MALDI–TOF-MS in the linear mode.

Electrophysiological measurement of g_j (junctional conductance)

Parental N2a cells were grown to 80 % confluence in a 12-well format and were transiently transfected with 1 μg of the pTracer-Cx50 cDNA plasmid for 4 h, lightly trypsinized and placed in 35-mm-diameter culture dishes overnight. Dual whole-cell patch-clamp experiments were performed the next day on GFP (green fluorescent protein)-expressing cell pairs to measure g_j, and the sensitivity to [Ca²⁺]_i was determined by permeabilizing the cells with 1 μM ionomycin and perfusing with ± 1.8 mM external CaCl₂ saline solutions. A +20 mV V_j (transjunctional voltage) pulse was applied every 15 s to monitor g_j during bath perfusion with nominally 0 or 1.8 mM [Ca²⁺]_o (external Ca²⁺ concentration) saline solutions. The g_j was normalized to the initial value (G_j = g_j/g_{j, initial}) for each Cx50 cell pair, and the results were pooled for each dataset. To examine the effect of CaM inhibitor or the candidate CaM-binding peptide on g_j, Cx50-N2a cells were pre-treated with 1 μM CDZ (calmidazolium) (Calbiochem) for 10–15 min or 1 μM of the synthetic Cx50 mimetic peptides (Cx50p^141–166 or SCx50p) were added to both whole-cell patch pipettes.

NMR spectroscopy

NMR experiments were performed with a Varian INOVA 600 MHz NMR spectrometer at 37 °C. (¹H,¹⁵N)-HSQC (heteronuclear single-quantum coherence) data were collected with 2048 complex data points and a spectral width of ~ 14 p.p.m. in the ¹H dimension, and 128 increments and a spectral width of 33 p.p.m. in the ¹⁵N dimension. A sample of 0.26 mM uniformly ¹⁵N-labelled CaM in buffer consisting of 5 mM Mes, 10 mM Bis-Tris (pH 6.5), 5 mM DTT (dithiothreitol), 0.1 mM sodium azide, 10 mM CaCl₂ and 10 % ²H₂O was titrated with peptide stock solutions (~ 5 mM) of the peptides Cx50p^141–166 or SCx50p in the same buffer. NMR data were processed using NMRPipe [24] and analysed using Sparky 3 (T.D. Goddard and D.G. Kneller, University of California, San Francisco).

The pulse-field gradient diffusion NMR spectra were collected with a modified PG-SLED pulse sequence [25] on a 600 MHz Varian INOVA spectrometer. In each FID (free induction decay) were contained 8k complex data points with a spectral width of ~ 13 p.p.m. The pulse-field gradient level (G) was arrayed from ~ 0.2 to ~ 31.0 Gauss/cm with a pulse gradient time (δ) of 5 ms and a diffusion time (Δ) of 112.5 ms. The data were processed using FELIX (Accelrys). The relationship between the NMR signal intensity (A) and the diffusion constant (D) follows eqn (1):

$$A=A_0\text{exp}[-(\gamma \delta G^2)(\mathrm{\Delta }-\delta /3)D]$$ (1)

where A₀ is the signal intensity when pulse gradient is not used and γ is the gyromagnetic ratio of the proton. In the data processing, the signal intensity as a function of pulse-field gradient level was fitted with eqn (2) using KaleidaGraph 3.5 (Synergy). The values discussed in the present paper were all determined by fitting with a linear correlation coefficient ≥0.999.

$$A=A_0\text{exp}(-C{G}^2)$$ (2)

where C is a constant [ = γδ(Δ − δ/3)D], the diffusion constant and the hydrodynamic radii R were obtained by comparing the C values of different molecules measured under identical conditions using eqns (3) and (4):

$$D/D_0=C/C_0$$ (3)

$$R/R_0=C_0/C$$ (4)

where D₀, R₀ and C₀ are the diffusion constant, hydrodynamic radius and the measured C value of lysozyme. The hydrodynamic radius and diffusion constant of lysozyme have been reported previously [26]. The gradients were calibrated at 25 °C on the residual ¹H signal in a sample of 99.9 % ²H₂O, using the published value of (1.902 ± 0.002) × 10⁻⁹ m² · s⁻¹ for the self-diffusion coefficient of ¹H²HO at 25 °C [27].

NMR samples contained 0.23 mM CaM in 10 % ²H₂O, 100 mM KCl and 10 mM imidazole at pH 6.5, with 10 mM Ca²⁺ for Ca²⁺-bound. The intensities of the protein signals were integrated from the methylene and methyl region of ~ 2 p.p.m. spectral width ( −0.2–1.8 p.p.m.). The integrated regions for each species were carefully selected to avoid or reduce interferences from buffer signals. The integrated intensities were normalized further to minimize the experimental errors from phase adjustment and baseline correction during the processing.

Chemical-shift perturbations (Δ) of the (¹H,¹⁵N)-HSQC spectra with and without peptide were calculated using both ¹H and ¹⁵N chemical shifts (δ) as shown in eqn (5):

$$\mathrm{\Delta }\delta =\sqrt{\frac{{(\mathrm{\Delta }\delta {}^1\text{H})}^2+{\left(\frac{\mathrm{\Delta }\delta {}^{15}\text{N}}{5}\right)}^2}{2}}$$ (5)

Fluorescence measurements

Steady-state fluorescence spectra were recorded using a QM1 fluorescence spectrophotometer (PTI) in a 1-cm-pathlength cell with a xenon short-arc lamp at 25 °C. The fluorescence emission spectra were acquired between 400 and 600 nm with an excitation wavelength at 335 nm for D-CaM. The slit widths were set at 4 nm for excitation and 8 nm for emission. A solution (0.8 ml) containing 0.25–2 μM D-CaM in 10 mM Tris/HCl (pH 7.4) and 100 mM KCl, with 5 mM Ca²⁺ or 5 mM EGTA was titrated by gradually adding 5–10 μl aliquots of the peptide stock solution (0.25–0.8 mM) in the same buffer. The binding constant of the synthetic peptide to modified CaM was obtained with a 1:1 binding model by fitting normalized florescence data as described previously [11,12].

The fluorescence anisotropy of D-CaM before and after addition of Cx50p^141–166 was measured with excitation at 335 nm and emission at 500 nm. An integration time of at least 20 s (20 points/s) was used to record the fluorescence signal, and all measurements were repeated at least three times by following established protocols [28,29].

Owing to domain-specific distribution of phenylalanine and tyrosine residues in CaM, the equilibrium Ca²⁺-binding constants for CaM were determined by monitoring fluorescence of intrinsic phenylalanine (λ_ex = 250 nm; λ_em = 280 nm;) for N-domain or tyrosine (λ_ex = 277 nm; λ_em = 320) for C-domain at 25 °C as described previously [30]. For N- or C-domain Ca²⁺-binding constants, 5–8 μM CaM was titrated with 5–10 μl aliquots of 50 or 15 mM Ca²⁺ stock solution in Ca²⁺ equilibrium buffer with 50 mM Hepes (pH 7.4), 100 mM KCl, 5 mM NTA (nitrilotriacetic acid) and 0.05 mM EGTA. The same method was applied for the 1:1 CaM/peptide mixture. Free Ca²⁺ at each titration point was determined with the Ca²⁺ dye Oregon Green 488 BAPTA-5N [5-nitro-1,2-bis-(o-aminophenoxy)ethane-N,N,N′,N′-tetra-acetic acid] (0.2 μM; K_d = −21.7 ± 2.5 μM [12]) at an emission of 520 nm with excitation of 495 nm. The free concentration was obtained using eqn (6):

$${[{\text{Ca}}^{2+}]}_{\text{free}}=K_{\text{d}}\frac{F-F_{min}}{F_{max}-F}$$ (6)

where F is the fluorescence intensity of the dye at each titration point, [Ca²⁺]_free is the concentration of free ionized calcium in solution, and F_min and F_max represent the dye fluorescence intensities of Ca²⁺-free form and Ca²⁺-saturated dye respectively. The Ca²⁺-binding affinity was obtained by fitting to the non-linear Hill equation:

$$f=\frac{{[{\text{Ca}}^{2+}]}_{\text{free}}^h}{K_{\text{d}}+{[{\text{Ca}}^{2+}]}_{\text{free}}^h}$$ (7)

where f is the fractional change of intrinsic fluorescence intensity, K_d is the Ca²⁺ dissociation constant, and h is the Hill coefficient.

CD spectroscopy

CD spectra were recorded in the far-UV (190–260 nm) range on a Jasco-810 spectropolarimeter at ambient temperature (25 °C) using a 0.1-cm-pathlength quartz cuvette. The measurements of CaM and the CaM–Cx50p^141–166 complex (10 μM) were made in 10 mM Tris/HCl (pH 7.5) and 100 mM KCl with 5 mM CaCl₂ or 5 mM EGTA. All spectra presented were averaged for at least 15 scans, and the background signal from the buffer was removed from the sample signals. The far-UV CD spectra of the peptide in different percentages of TFE (trifluoroethanol) were obtained using a 20 μM concentration of Cx50p^141–166 in the same buffer. The secondary structures of the peptides were calculated with the online secondary-structure prediction server DICHROWEB [31]. The α-helical content of peptides was predicted using the Agadir algorithm [32–34].

Mass spectrometry

The MALDI–TOF-MS analysis was carried out on an Applied Biosystems 4800 plus MALDI–TOF/TOF analyser mass spectrometer. The data were acquired in a linear positive mode with SA (sinapinic acid) as matrix for CaM (50 μM) and the CaM–Cx50p^141–166 complex. The molecular mass of Cx50p^141–166 was also confirmed by MALDI–TOF-MS with CHCA (α-cyano-4-hydroxycinnamic acid) as matrix. CaM (50 μM) and Cx50p^141–166 (50 μM) were mixed in 50 mM Tris/HCl (pH 7.5) and 100 mM KCl, and a 1 μl mixture was added with 10 μl of saturated SA solution and then dried on the MALDI plate for the measurement.

RESULTS

A putative CaMBD in the single cytoplasmic loop of Cx50

The CaM-binding site predication server, Calmodulin Target Database, was used to identify potential CaMBDs within Cx50, by taking into account the hydropathy, α-helical propensity, residue mass, residue charge, hydrophobic residue content and helical class [14]. The highest predictive score was assigned to a stretch of sequences (residues 141–166; Figure 1) in the C-terminal portion of the intracellular loop of Cx50. Sequence alignments with previously characterized CaMBDs in rodent Cx43 [11] and sheep Cx44 (human Cx46) [12], and secondary-structure prediction, identified common features among the three α-family Cxs that included: (i) the spacing of hydrophobic residues with bulky side chains, often used as anchor in CaM target complexes [35–37], followed a 1–5–10 pattern; (ii) all of the domains exhibited strict conservation in positions with positively charged residues, which help to drive the formation of CaM target complexes via electrostatic interactions [38]; and (iii) the CaMBDs were predicted to have a predominantly α-helical structure. Helical wheel analysis further suggests the clustering of hydrophobic residues on one side and the frequent occurrence of charged residues on the opposite side. A peptide corresponding to the predicted CaMBD was subsequently synthesized for the functional and biophysical studies described below.

Figure 1. Cxs membrane topology and the putative CaM-binding sites.

The α-class of Cxs are composed of four transmembrane segments, two extracellular loops, one cytoplasmic loop, a short N-terminus and a much longer C-terminal tail. The predicted CaM-binding sites are located in the second half of the intracellular loop between TM2 and TM3. The numeric score (1–9) represents the probability of an accurate prediction of high-affinity CaM-binding sites. The CaM-binding sequences identified in the α-class Cxs are similar to those of CaMKII. All of the aligned sequences fit the 1–5–10 CaM-binding mode subclass, where each number represents the presence of a hydrophobic residue. m, mouse; h, human; s, sheep; r, rat; #, hydrophobic residues (highlighted in grey); B, basic residues (underscored).

Ca²⁺-dependent uncoupling of Cx50 gap junctions requires CaM

To determine whether Cx50 gap junctions like those comprising Cx43 [11,39] and Cx46 [12] are regulated by [Ca²⁺]_i, we first performed dual whole-cell patch-clamp experiments in Cx50-N2a cells. Perfusion with saline solution containing 1 μM ionomycin+1.8 mM [Ca²⁺]_o induced a time-dependent reduction in G_j (Figure 2A and Table 1); we have shown previously that such treatment effects an elevation of [Ca²⁺]_i in HeLa cells [40]. Omission of 1.8 mM CaCl₂ from the bath saline prevented the Cx50 gap junction uncoupling response in the presence of 1 μM ionomycin.

Figure 2. Ca²⁺- and CaM-dependence of Cx50 gap junction uncoupling.

(A) Cx50-N2a cell pair superfusion with 1 μM ionomycin bath saline with or without 1.8 mM CaCl₂ demonstrates a time-dependent uncoupling of Cx50 gap junctions that requires 1.8 mM [Ca²⁺]_o. (B) Pre-treatment with 2 mM CDZ for 10–15 min before patch-clamp analysis completely prevented the 1.8 mM [Ca²⁺]_o-dependent uncoupling response. (C) Inclusion of 1 μM Cx50p^141–166 in both whole-cell patch pipettes prevented 80 % of the Ca²⁺/CaM-dependent decline in Cx50 G_j. In contrast, experiments performed with SCx50p prevented only 20 % of the 1.8 mM [Ca²⁺]_o–CaM-dependent reduction in Cx50 G_j. The average ± S.E.M. initial g_j values for all experimental groups are displayed in Table 1.

Table 1.

Electrophysiological measurement

Experiment	Initial gj (nS)	N
1.8 mM CaCl2	9.90 ± 2.90	5
0 mM CaCl2	14.74 ± 6.51	4
2 mM CDZ	16.79 ± 2.95	4
Cx50p141–166	7.05 ± 2.53	7
SCx50p	3.36 ± 1.80	6

g_j = Ij/V j (Ohm’s law, junctional current divided by transjunctional voltage). N = number of cell pairs recorded from each dataset.

To test whether the Ca²⁺-dependent decrease in Cx50 G_j produced by addition of 1 μM ionomycin+1.8 mM [Ca²⁺]_o required CaM, Cx50-N2a cell cultures were acutely (≤15 min) treated with CDZ (2 μM), a CaM inhibitor with a K_d of ~ 1 μM [41]. Pre-treatment with CDZ prevented the Ca²⁺-dependent decrease in G_j (Figure 2B).

A more specific test for a role of Cx50 CaMBD in mediating the Ca²⁺-dependent inhibition of this Cx was to add the 26-mer Cx50 mimetic peptide corresponding to residues 141–166 of the cytoplasmic loop to the Cx50-transfected N2A cells. A scrambled peptide containing a randomized sequence of the same amino acids was also utilized as a control. The addition of Cx50p^141–166, but not the scrambled peptide, prevented the decline in G_j (Figure 2C). Overall, these electrophysiological data support the hypothesis that an elevation of [Ca²⁺]_i can uncouple Cx50 gap junctions in a CaM-dependent manner mediated by the Cx50 141–166 domain.

To examine the molecular basis for the Ca²⁺-dependent decrease in Cx50 G_j, gap junction channel recordings were analysed in two poorly coupled cell pairs. Both experiments exhibited a maximum of three open channels (N) and their cumulative open probability (N · P_o) declined from 1.21 open channels per unit time to 0 upon exposure to 1 μM ionomycin and 1.8 mM [Ca²⁺]_o saline (Figure 3A). Unitary Cx50 gap junction channel currents remained constant (314 ± 18 pS, mean ± S.D., n = 14) during the decline in P_o, indicative of a conformational gating type behaviour (Figure 3B).

Figure 3. Mechanistic basis for Ca²⁺/CaM-dependent Cx50 gap junction uncoupling.

(A) The product of the number of open channels (N) and open probability (P _o) from two poorly coupled Cx50-N2a cell pairs is plotted as a function of time and 1 μM ionomycin +1.8 mM [Ca²⁺]_o superfusion. N · P_o declined from an average control value of 1.21 open channels to 0 within 2 min of ionomycin/CaCl₂ saline superfusion. (B) Cx50 gap junction channel current recordings from one experiment are shown for the time points indicated by the circles in (A). The number and duration of the open channels declined progressively to 0 without any apparent reduction in the single-channel conductance (314 pS), indicative of closure of a Ca²⁺–CaM-dependent gate without a block of the ion-permeation pathway.

Ca²⁺-dependent specific interaction between Cx50 CaMBD and CaM revealed by NMR

After confirming the requirement for CaM in the Ca²⁺-dependent regulation of the Cx50 gap junctions, we further asked the question whether or not there is a direct interaction between the predicted CaMBD (or Cx50p^141–166) and CaM. The formation of the CaM–Cx50p^141–166 complex was first confirmed by MALDI–TOF-MS (see Supplementary Figure S1 at http://www.BiochemJ.org/bj/435/bj4350711add.htm). In the presence of Ca²⁺, a distinct peak representing the CaM–Cx50p^141–166 complex (theoretical molecular mass 19.93 kDa; experimental molecular mass 20.02 kDa) was detected.

To identify residue-specific changes in the conformation of CaM during complex formation with Cx50p^141–166, we carried out (¹H,¹⁵N)-HSQC NMR experiments. Unlabelled Cx50p^141–166 stock solution was gradually titrated into ¹⁵N-labelled CaM. Under Ca²⁺-loaded conditions, the HSQC spectrum of Ca²⁺-saturated CaM (holo-CaM) underwent significant changes on addition of 1 mol. eq. of Cx50p^141–166 (Figure 4A). In contrast, SCx50p, which is predicted to have no binding affinity for CaM, was similarly added to ¹⁵N-labelled holo-CaM. This scrambled peptide failed to induce chemical-shift changes in CaM (Figure 4B), therefore ruling out the possibility of non-specific interactions between peptide and CaM.

Figure 4. Monitoring the interaction between CaM and Cx50p^141–166 by (¹H,¹⁵N)-HSQC spectroscopy.

(A) An overlay of HSQC spectra of holo-CaM (black) with the spectrum of the holo-CaM–Cx50p^141–166 (grey). (B) An overlay of the HSQC spectra of holo-CaM (black) with the spectrum of the holo-CaM–SCx50p (grey). (C) Chemical-shift perturbation in CaM induced by addition of 2-fold molar excess of Cx50p^141–166. The weight-average chemical-shift change (Δδ) was calculated using eqn (5). Residues with Δδ> 0.1 p.p.m. were mapped to the three-dimensional structure of holo-CaM (PDB code 3CLN). (D) The chemical-shift change of Gly³³ during titration of holo-CaM with Cx50p^141–166. The disappearance of the peak (free form) was accompanied by the appearance of the corresponding peak (bound form) at a new position. (E) Structural basis of the difference in directionality of chemical shift change. The structure of the CaM–CaMKII complex (PDB code 1CDM) is shown with the peptide (inside) and the CaM residues (outside). Leu¹¹⁶ and Met¹⁴⁵ of CaM are within 5 Å of Lys²⁹⁸p and Lys³⁰⁰p of the peptide respectively. These peptide positions correspond to Lys¹⁴⁶p and Arg¹⁴⁸p for Cx43, and Arg¹⁴⁹p and Glu¹⁵¹p for Cx50, as can be seen in the superposition of their sequences. In Cx50, Glu¹⁵¹p may be pulled away from Met¹⁴⁵ by Lys¹⁴⁷p. In contrast, in Cx43, Arg¹⁴⁸p may be pushed towards Met¹⁴⁵ by Lys¹⁴⁴p. These differences in peptide sequences can cause changes of chemical shifts in different directions.

By analysing the NMR chemical shifts of CaM–peptide complexes deposited in the Biological Magnetic Resonance Bank (BMRB codes 1634, 4270, 5480, 5770, 5893, 5896, 15470 and 16465), we identified some significant changes in chemical shifts [δ(CaM) − δ(complex) > 1 S.D.] that are common to all complexes such as the ¹H shifts of Ala⁵⁷ and Met⁷¹, and ¹⁵N shift of Ala¹²⁸. In the classical CaM–CaMK (Ca²⁺/CaM-dependent protein kinase) IIα peptide structure with 1–10 binding mode, Met⁷¹ and Ala¹²⁸ are in the hydrophobic pockets of CaM and are within 5 Å (1 Å = 0.1 nm) of the peptide. Ala⁵⁷ occupies the second position of the second Ca²⁺-binding loop, suggesting that peptide binding leads to conformational change in this loop. In the CaM–Cx50p^141–166 complex, Ala⁵⁷ and Met⁷¹ have δ_H values of 0.29 and 0.05 p.p.m. respectively, whereas Ala¹²⁸ has a δ_N of 0.21 p.p.m., consistent with Cx50p^141–166 bound to Ca²⁺–CaM.

Upon the formation of a Ca²⁺/CaM–Cx50p^141–166 complex, the most significant perturbations, with weighted average chemical shift changes greater than 0.1 p.p.m., were observed in the N-lobe (e.g. Ser¹⁷, Thr²⁶, Thr²⁹, Glu³¹, Leu³⁹ and Ala⁵⁷), the C-lobe (e.g. Glu⁸², Asp⁹³, Leu¹⁰⁵, Arg¹⁰⁶, Ala¹⁴⁷ and Lys¹⁴⁸) and the linker region (e.g. Asp⁶⁴ and Phe⁶⁸), indicating that Cx50p^141–166 causes a global conformation changes in CaM. A detailed residue–residue plot of chemical-shift perturbation is shown in Figure 4(C). The overall change of the N-lobe residues (2.74 p.p.m.) was greater than in the C-lobe residues (1.78 p.p.m.). During the titration, we observed the simultaneous appearance of two peaks representing the same residue. For example, a progressive disappearance of the amide signal of Gly³³ representing unbound CaM was accompanied by the concomitant emergence of a new set of peaks from the CaM–Cx50p^141–166 complex (Figure 4D). Such a slow exchange process, which occurs when the exchange rate of the bound and unbound states is smaller than the amide frequency difference, often heralds a high affinity protein–protein or protein–peptide association (i.e. interactions with submicromolar affinities) [12,42].

Probing the CaM–peptide complex state by diffusion NMR

Pulse-field gradient NMR was performed to determine the diffusion constant of the CaM–Cx50p^141–166 complex, enabling us to assess the overall size of this complex in solution. The diffusion constant was determined to be (12.4 ± 0.4) × 10⁷ cm/s (Figure 5). By assuming a spherical shape of the complex, the calculated size of the Ca²⁺/CaM–Cx50p^141–166 complex was determined to be 17.3 ± 0.6 Å, a value comparable with the hydrodynamic radii of CaM in complex with well-known targets such as smMLCK (smooth muscle myosin light chain kinase) (17.9–21.8 Å), PDE (phosphodiesterase) (18.8–22.3Å) and CaMKI (21.2 Å) [43]. The hydrodynamic radius of CaM (22.6 ± 0.6 Å) was decreased by 23 % on formation of the CaM–Cx50p^141–166 complex. This size conforms to a collapsed structure of the complex that involves the unwinding of the central helix within CaM to embrace the target peptide.

Figure 5. Hydrodynamic radii of CaM–Cx50p^141–166 complex determination by pulse-field gradient NMR.

The NMR signal decay of holo-CaM (○) and holo-CaM–Cx50p^141–166 complex (●) as a function of field strength. The calculated hydrodynamic radii of the CaM and complexes are indicated on the top.

Revealing CaM–peptide interactions with CD spectroscopy

CD spectra also confirmed the structural and conformational changes during the formation of the CaM–Cx50p^141–166 complex. As indicated in Figure 6(A), the addition of Cx50p^141–166 to holo-CaM at a 1:1 molar ratio resulted in a ~ 6 %increase in the negative ellipticity at 222 nm. The hydrophobic environment in the peptide-binding pocket of holo-CaM could be the reason for the increase in the negative ellipticity attributed to the peptide itself. In contrast, a 5 % decrease at 222 nm and a 20 % decrease at 208 nm in the CD signal intensity were observed after the addition of Cx50p^141–166 to apo-CaM. Further exploring the interaction of apo-CaM with Cx50p^141–166 using NMR was not possible because of limited solubility of the peptide in 5 mM EGTA (see Supplementary Figure S2 at http://www.BiochemJ.org/bj/435/bj4350711add.htm).

Figure 6. CD studies of the interaction between CaM and Cx50p^{141 – 166}.

(A) Far-UV CD spectra of CaM in the presence of 5 mM EGTA (○, apo-CaM) or CaCl₂ (□, holo-CaM), and a CaM–Cx50p^141–166 (1:1) mixture with 5 mM EGTA (●, apo-CaM–Cx50p^141–166) or CaCl₂ (■, holo-CaM–Cx50p^141–166). (B) Far-UV spectra of the synthetic peptide Cx50p^141–166 with addition of TFE (0–80 %). The inset shows calculated α-helical content as a function of TFE concentration for the peptides Cx50p^141–166 (■), Cx44p^129–150 (▲) and Cx43p^136–158 (●). (C) Far-UV spectra of the synthetic peptide Cx50p^141–166 with (●) or without (○) addition of CaM.

We examined further the secondary structure of Cx50p^141–166 under different solvent conditions. TFE is known to induce α-helical formation of peptides, and the helical content in the presence of TFE usually reflects its helical propensity [44]. Cx50p^141–166 becomes predominantly helical (>90 %) at TFE ≥30 % (v/v) (Figure 6B, inset), suggesting that the Cx50 CaM-binding sequence readily adopts an α-helical configuration. Cx50, Cx44 and Cx43 exhibit 94, 55 and 33 % helical conformation respectively in the presence of 30 %TFE (Table 2).

Table 2.

Binding affinities to CaM and α-helicity of the Cx peptides

Peptide	Kd (nM)		α-Helicity (%)
	5 mM Ca2+	5 mM EGTA	Prediction	Far-UV CD measurement
Cx50p141–166	4.9 ± 0.6	>8000	1.11	94
Cx44p129–150	49 ± 3.0	>5000	1.20	55
Cx43p136–158	860 ± 20	nd	0.52	33

All of the experiments were conducted three times (n = 3). nd, not detectable. K_d measurements were made of D-CaM fluorescence in 50 mM Tris/HCl (pH 7.5) and 100 mM KCl. The prediction of the helical content of peptides used the Agadir algorithm. Far-UV CD measurements were conducted using 30 % TFE in 10 mM Tris/HCl (pH 7.5) and 100 mM KCl.

The peptide in the absence of CaM has only 7 % α-helix and 72 % random coil. The addition of peptide into CaM resulted in the addition of helicity as shown in Figure 6(A). Figure 6(C) shows that our difference CD spectrum of the CaM in the presence and absence of the peptide exhibits two negative minima at 208 and 222 nm. To understand the origin of the additional helicity due to the formation of the CaM–Cx50p^141–166 complex, we analysed the helical contents in the X-ray structure of Ca²⁺-loaded calmodulin alone (PDB code 3CLN) and the complex structure of CaM–CaMKII peptide (PDB code 1CDM) since our chemical-shift studies suggest that CaM binds to Cx50p^141–166 in a similar binding mode as to the peptide of CaMKII (Figure 4E). CaM does not exhibit any additional change in its helical content upon formation of the complex with a helicity of ~ 62 %. Thus the complex-induced helicity is attributed to the conversion of peptide to be ~ 20 % helical.

Determination of the CaM–peptide binding affinity by fluorescence spectroscopy

D-CaM has frequently been used to determine the binding affinity of the CaM–peptide interaction due to its sensitivity to the changes in the surrounding chemical environment and its ease of preparation [22]. The dansyl moiety has an emission maximum at ~ 500–510 nm, a range in which the signals from intrinsic aromatic residues are negligible. We first examined the dansyl fluorescence anisotropy changes to confirm the formation of complex in the low-micromolar range. Fluorescence anisotropy is dependent on the rotational correlation time of fluorophore in the sample, and is often used to reflect the hydrodynamic properties of biomolecules [45]. The increase in anisotropy often arises from slower tumbling of the fluorophore and thus reports the formation of a larger complex. Indeed, upon addition of 1 mol. eq. of Cx50p^141–166, the anisotropy of the dansyl moiety within D-CaM increased from 0.083 to 0.125, suggesting the association of Cx50p^141–166 with Ca²⁺–CaM (see Supplementary Figure S3 at http://www.BiochemJ.org/bj/435/bj4350711add.htm).

We then carried out the titration of peptide with D-CaM by monitoring the fluorescence emission between 400 and 600 nm. In the presence of Ca²⁺, D-CaM showed a fluorescence maximum at 500 nm. With the addition of Cx50p^141–166, the dansyl fluorescence emission maximum blue-shifted to 474 nm with a concomitant enhancement of its fluorescence intensity (Figure 7A), implying that the dansyl group entered a more hydrophobic environment upon complex formation with peptide. In contrast, in the presence of EGTA (Figure 7B), although the dansyl fluorescence intensity also increased, the fluorescence maxima changed less than in the presence of Ca²⁺. The data confirm the CD signal change on the addition of peptide to apo-CaM. By fitting the titration curve with a 1:1 binding mode, we obtained a K_d of 4.9 ± 0.6 nM for the interaction between Ca²⁺–CaM and Cx50p^141–166 (Table 2).

Figure 7. Interaction of Cx50p^141–166 with D-CaM monitored by steady-state fluorescence.

(A) The titration curve of D-CaM (2 μM) with Cx50p^141–166 in the presence of 5 mM Ca²⁺ in a buffer consisting of 50 mM Tris/HCl (pH 7.5) and 100 mM KCl. The inset shows the fluorescence spectrum of D-CaM in the absence (broken line) or the presence (continuous line) of an equivalent molar concentration of Cx50p^141–166. (B) Titration curve of D-CaM (2 μM) with Cx50p^141–166 in the presence of 5 mM EGTA in a buffer consisting of 50 mM Tris/HCl (pH 7.5) and 100 mM KCl. The inset shows the fluorescence spectrum of D-CaM in the absence (broken line) or the presence (continuous line) of an equivalent molar concentration of Cx50p^141–166. (C) pH-dependence of Cx50p^141–166 and Cx44p^129–150 binding to CaM. The binding affinities were derived from the peptide titration curves of D-CaM at various pH values. All experiments were conducted in triplicate.

To investigate further the possible role of pH on CaM–peptide association, the effect of varying pH on the binding affinity of CaM for Cx50p^141–166 was also examined. As shown in Figure 7(C), the binding affinity (represented as − log K_d) exhibited a pH-dependent change between pH 5.0 and 10.0. The highest binding affinity was obtained at pH 6.5. The observed pH-dependence of the interaction between CaM and Cx50p^141–166 indicates that electrostatic interactions are likely to be the main force driving CaM–peptide complex formation.

Binding of Cx50p^141–166 expands the [Ca²⁺]_i-sensing range of CaM

We next performed the Ca²⁺ equilibrium titration of CaM to obtain macroscopic Ca²⁺-binding constants for both the N- and C-lobes of CaM by monitoring domain-specific fluorescence changes as described previously [29,30]. The decrease in phenylalanine fluorescence (λ_ex = 250 nm; λ_em = 280 nm) and the increase in tyrosine fluorescence (λ_ex = 277 nm; λ_em = 320 nm) report Ca²⁺ binding to the N-lobe and the C-lobe of CaM respectively. This allows us to examine the effect of peptide binding on the domain-specific Ca²⁺-sensing capability of CaM. As shown in Figure 8 and summarized in Table 3, the addition of Cx50p^141–166 led to a rightward shift of the titration curve, representing the Ca²⁺ binding to the N-lobe, and concomitantly caused a leftward shift of the titration curve, reflecting the Ca²⁺ binding to the C-lobe of CaM. Specifically, the Ca²⁺ affinity of the N-lobe of Cx50p^141–166-bound CaM decreased 2-fold, whereas the Ca²⁺ affinity of the C-lobe increased by ~ 20 %. As a result of Cx50p^141–166 binding, CaM responded to subtle changes in [Ca²⁺]_i over a broader range of calcium concentrations.

Figure 8. Domain-specific equilibrium Ca²⁺ titration of CaM (○) and CaM in complex with Cx50p^141–166 (●).

(A) The intrinsic phenylalanine fluorescence (λ_ex = 250 nm; λ_em = 280 nm) was monitored to report the equilibrium Ca²⁺-binding constants of the N-lobe of CaM. (B) The intrinsic tyrosine fluorescence (λ_ex = 277 nm; λ_em = 320 nm) was monitored to report the equilibrium Ca²⁺-binding constants of the C-lobe of CaM. The free ionized Ca²⁺ concentration was measured using the Ca²⁺ indicator dye Oregon Green 488 BAPTA-5N. All experiments were conducted in triplicate in 50 mM Hepes (pH 7.5), 100 mM KCl, 5 mM NTA and 0.5 mM EGTA.

Table 3.

Effects of Cx peptides binding on the metal-binding properties of CaM

Peptide	N-domain (sites I and II)		C-domain (sites III and IV)
	Kd (μM)	h	Kd (μM)	h
None	12.0 ± 0.2	1.4 ± 0.1	3.78 ± 0.03	2.14 ± 0.03
Cx50p141–166	22.7 ± 0.6	0.8 ± 0.1	3.18 ± 0.03	2.74 ± 0.07
Cx44p129–150	11.6 ± 0.1	1.5 ± 0.1	0.93 ± 0.02	2.20 ± 0.10
Cx43p136–158	14.5 ± 0.1	1.6 ± 0.2	1.16 ± 0.02	2.10 ± 0.10

All of the experiments were conducted three times (n = 3). K_d and the Hill coefficient (h) were obtained by fitting the titration curve to eqn (7). Phenylalanine fluorescence (λ_ex = 250 nm; λ_em = 280 nm) reports the Ca²⁺ binding to the N-domain of CaM. Tyrosine fluorescence (λ_ex = 277 nm; λ_em = 320 nm) reports the Ca²⁺ binding to the C-domain of CaM.

DISCUSSION

Regulation of Cxs by intracellular Ca²⁺ and CaM

The results of the present study demonstrate that intracellular calcium is important in the regulation of Cx50 gap junctions [40]. Furthermore, using the CaM inhibitor CDZ as well as a peptide fragment encompassing the predicted CaM-binding region, we were able to demonstrate that this Ca²⁺-dependent inhibition of Cx50 gap junctions is mediated by the ubiquitous intracellular Ca²⁺ receptor CaM. Thus CaM appears to mediate the Ca²⁺-dependent regulation of all three major α-Cxs in the lens, i.e. Cx43, Cx46 and Cx50. Our results also demonstrate that this inhibition of Cx50 gap junction channels by Ca²⁺ is due to a decrease in channel open time probability as opposed to a reduction in single-channel conductance (Figure 3A). These records provide the first direct evidence for the gated closure of individual Cx gap junction channels by Ca²⁺–CaM. The average observed g_j value of 314 pS is indicative of Cx50, since it is at least double the observed g_j for Cx40, Cx43 or Cx45 gap junctions in our Cx50-N2a cell cultures and consistent with the reported 210–290 pS values from reconstituted lens fibre cell membranes or exogenously expressed Cx50 gap junction channels ([46–50], and Q. Xu, Y. Chen, J.J. Yang and R.D. Veenstra, unpublished work).

The possible role of CaM in the chemical gating of the β-family Cxs was first proposed by Peracchia and Wang [51]. CaM co-localizes with Cx32 [9] and directly gates Cx32-containing gap junctions [52]. Török et al. [9] reported that fluorescently labelled CaM derivatives bind to synthetic peptides spanning most of the cytoplasmic sequences from two regions of Cx32: an N-terminal sequence comprising residues 1–21 (K_d = 27 nM) or residues 1–19 (K_d = 1.1 μM); and a C-terminal sequence comprising residues 216–230 (K_d = 2.1 μM) or residues 208–226 (K_d = 3.5 μM). Both exhibit Ca²⁺-dependent CaM-binding properties [9,10]. They also identified two distinct CaM-binding amino acid sequences in Cx32 within the N- and C-lobes of CaM showing separate functions, suggesting trans-domain or trans-subunit bridging by CaM as a possible mechanism of gap junction gating [52]. In addition, Ahmad et al. [53] using in vitro translation approaches have shown that oligomerization of Cx32 is CaM-dependent, since CaM interacts with Cxs at an early stage of gap junction assembly [8,9]. The C-terminal tail of the CaM-binding site of Cx32 is likely to be involved in the assembly of Cxs and is Ca²⁺-dependent [8]. Furthermore, Blodow et al. [54] reported that CaM antagonists suppress gap junction coupling of Cx26 in isolated Hensen cells of the guinea pig cochlea. Taken together, [Ca²⁺]_i and CaM play vital roles in regulating gap junctions mediated by both the α- and β-family of Cxs.

Comparison of CaM-binding affinity with the α-family Cxs

Using high-resolution NMR, CD and fluorescence studies, we have shown using a peptide model that CaM interacts with the predicted cytosolic CaMBD of Cx50. Our current studies, as well as previous studies, have clearly shown that CaM exhibits a Ca²⁺-dependent interaction of all three peptide fragments encompassing the intracellular loop regions of all three lens Cxs. Fluorescence spectroscopy revealed conformational changes of both the peptide and CaM following formation of the CaM–peptide complex. NMR studies demonstrated that the peptide binds to CaM with a 1:1 stoichiometry. It also indicated that the peptide induces structural changes in both the N- and C-terminal domains, as well as in the linker region of CaM and the binding of Cx peptides to CaM reflects a classical embracing mode of interaction. Overall, for these three Cxs, binding of the peptide to CaM decreases the apparent K_d of Ca²⁺ for CaM, and the Hill coefficient increased. In addition, Cx43 exhibits weaker affinity for CaM than does either Cx50 or Cx44.

Consistent with the CaM–Cx peptide binding affinity studies using fluorescence spectroscopy, our NMR chemical-shift analysis also supports a lower affinity for the CaM–Cx43p^136–158 complex. The addition of Cx43p^136–158 and Cx50p^141–166 results in similar changes in the CaM spectra as shown in Supplementary Figure S4 (at http://www.BiochemJ.org/bj/435/bj4350711add.htm). For most signals, the directionalities of the changes are the same, reflecting the same 1–5–10 binding mode of these peptides to CaM. However, we have observed the slow exchange process for Gly³³ in Cx50 shown in Figure 4(D). A similar scenario was observed for Thr²⁹, Ala⁵⁷ and Thr¹¹⁷ for the CaM–Cx44p^129–150 complex [12]. Such a slow exchange process occurs when the exchange rate of the bound and unbound states is smaller than the amide frequency difference, and often heralds a high-affinity protein–protein or protein–peptide association (i.e. interactions with submicromolar affinities) [42] that is consistent with the CaM-binding affinity to Cx50p^141–166 and Cx44p^129–150. In contrast, the NMR signals of CaM upon addition of Cx43p^136–158 exhibits a gradual change in chemical shifts characteristic of the fast exchange regime, suggesting a weaker binding affinity.

Implications for the factors contributing to the CaM-binding affinity to Cxs

There are several positively charged residues, such as Lys¹⁴⁷, Arg¹⁴⁹ and Arg¹⁵⁶, in Cx50 that are conserved in all α-family members. To identify the key factors responsible for the binding affinity of CaM to Cxs, we have determined the CaM-binding affinity of Cx50p^141–166 to CaM as a function of pH. The peptide-binding affinity is decreased at pH values lower than 5.5 and greater than 8.5. This trend is very similar to our previous results reported for Cx44 [12]. Both Cx44p^129–150 and Cx50p^141–166 behave similarly when pH is varied, but CaM binding to Cx50p^141–166 is ~ 1 order of magnitude stronger than to the Cx44p^129–150. Such a similar pH-dependence suggests that electrostatic interactions following the protonation of aspartate and glutamate and deprotonation of lysine are important for this complex formation.

The difference in CaM-binding affinity of Cx50p^141–166 and Cx44p^129–150 may originate from the intrinsic sequences encoded in the CaM-binding regions of different Cxs. As shown in Figure 1, there are several differences in the predicted regions especially with variations between conserved residue positions 1 and 5. Cx43 contains two flexible glycine residues instead of the helix-forming residues in Cx44 and Cx50. Such sequence variation reflects their ability to form α-helices that are important for their interaction with CaM [14]. As shown in Figure 6(B) and Table 2, Cx43p^136–158 has the lowest α-helix-formation capacity. The addition of TFE (>60 %) induces the highest α-helical content in the Cx50p^141–166 and Cx44p^129–150. In Figure 6(B), the α-helicity of Cx50p^141–166 is higher than that of Cx43p^136–158 and Cx44p^129–150 reported previously [11,12]. The calculated α-helical content of peptides in TFE exhibits the rank order Cx50p^141–166>Cx44p^129–150>Cx43p^136–158. Interestingly, negatively charged Glu¹⁵¹ located close to Arg¹⁴⁹ in Cx50 may stabilize the α-helical conformation of this peptide. In contrast, Cx43 has a positively charged arginine residue at the same location that may disable α-helical conformation. The α-helical content of the CaM-binding peptides observed is consistent with the trend for the CaM-binding affinity results from our fluorescence experiments (Table 2). Both Cx44 and Cx50p^141–166 also bind CaM in the absence of Ca²⁺, as revealed by fluorescence data. The CaM-binding affinities to Cx44 and Cx50 in EGTA are 100- and 1000-fold weaker than those in the presence of Ca²⁺ respectively. The apo-form of CaM does not interact with Cx43p^136–158, which has the least α-helical content. Therefore our results reveal that the α-helicity of the CaM-binding peptide is an important factor in predicting the CaM-binding affinity of these Cxs.

Our detailed NMR analysis has also revealed the differential interactions of CaM with Cx50p^141–166 and Cx43p^136–158. We observed that, in general, the N-lobe of CaM–Cx50p^141–166 has larger chemical-shift changes, whereas the C-lobe has smaller changes than those of CaM–Cx43p^136–158, although such interactions belong to a similar 1–5–10 binding mode. In addition, chemical shifts of residues such as Phe¹², Lys¹³, Lys⁹⁴, Leu¹¹⁶, Ile¹³⁰ and Met¹⁴⁵ of CaM–Cx50p^141–166 changed in different directions compared with those of CaM–Cx43p^136–158 (see Supplementary Figure S4). Among these residues, Lys⁹⁴ and Ile¹³⁰ are on the third and fourth Ca²⁺-binding loops respectively, and these loops could be flexible and adopt different conformations, leading to different directions of change in chemical shifts. In the structure of CaM–CaMKIIα (PDB code 1CDM), Leu¹¹⁶ interacts with Lys²⁹⁸, which corresponds to Lys¹⁴⁶ in Cx43 and Arg¹⁴⁹ in Cx50. Similarly, Met¹⁴⁵ interacts with Lys³⁰⁰, which corresponds to Arg¹⁴⁸ in Cx43 and Glu¹⁵¹ in Cx50 (Figure 4E). Such changes in directions of the chemical shifts of CaM are likely to be due to the intrinsic properties such as the α-helical propensity of the sequences in the CaM-binding regions of Cxs.

Structural and disease implications

The intracellular region regulated by CaM has been reported to be involved in several disease-related mutations. ODDD (oculodentodigital dysplasia) is caused by mutations in Cx43 such as G138R, G138S, G143S, G143D, K144E, V145G, M147T, R148G, R1481, T154A and T154N [55], overlapping with the CaM-binding region identified previously [11]. There have been no reports of point mutations in the CaMBD of Cx50 associated with cataract formation. However, Beahm and Hall [56] have reported that the Cx50 H161N mutant does not form detectable hemichannels, but forms gap junctions indistinguishable from wild-type. Interestingly, Toloue et al. [57] identified three nonfunctional mutations in Cx50, T157C, H161C and E169C, that failed to function homotypically or in heterotypic pairings with wild-type Cx50. The threonine residue at position 157 is highly conserved and this residue has been shown previously to be functionally essential for all Cxs tested, including Cx26, Cx43 and Cx50 [58]. Mutation of a conserved threonine residue in the third transmembrane helix of α- and β-Cxs creates a dominant-negative closed gap junction channel [58]. It is possible that such mutations may compromise the regulation of this Cx by CaM. The testing of such mutational effect on Cx50 function is currently underway in this laboratory.

The molecular scenario of CaM-mediated regulation of the α-class of gap junctions seems to be quite different from the β-class, probably due to the fact that the α-class Cxs lack the N-terminal glycine hinge and possess longer intracellular loops and C-termini than the β-class Cxs. Indeed, the CaM-targeting sequences in the α-class Cxs, such as rodent Cx43, sheep Cx44 (or human Cx46) and mouse Cx50, as revealed in the present study, are in the cytoplasmic loop connecting TM2 and TM3, in close proximity to the cytoplasmic loop–TM3 interface. The apparent juxtaposition of the Cx32 N-terminal CaMBD to TM1 and of the Cx43 cytoplasmic loop CaMBD to TM3 may facilitate the closing of the Cx channel pore since both TM1 and TM3 are suggested to be involved in forming the transmembrane pore while the N-terminus loops back to form a pore cytoplasmic vestibule.

The recently determined three-dimensional structure of the Cx26 gap junction channel at 3.5 Å resolution provides insights into the gating mechanism of the β-class of gap junctions at atomic resolution [59]. In this tsuzumi (Japanese drum)-shaped structure, the permeation pathway is mainly defined by the short NTH (N-terminal helix) (residues 2–10) that lines the pore funnel of gap junction channel and TM1 that constitutes the pore-lining helix. This structure suggests that the molecular identity of the channel ‘plug’ observed using electron microscopy [60] is likely to be the NTH. However, owing to the invisibility of both the cytoplasmic loop and the C-terminus in the crystal structure, both of which play important roles in the chemical gating of gap junctions [61], it is still too premature to explain the chemical gating mechanism with the current structure.

Given that the N-terminal CaMBD in Cx32 overlaps with the NTH that is believed to form the channel ‘plug’ in β-class gap junctions, it is conceivable that the Ca²⁺-dependent binding of CaM to NTH might function as ‘cork’ to block the channel by the complex or by inducing conformational changes in the NTH. However, since the NTH is lined interior to the channel to form a funnel, its accessibility to CaM in the assembled channel remains to be confirmed.

Despite the absence of X-ray structural information about the intracellular loop of Cxs [59], we have provided detailed structural information about the association of CaM with all three α-family Cx members. Our results provide the first direct evidence that CaM binds to a specific region of the lens gap junction Cxs Cx43, Cx44 and Cx50 in a Ca²⁺-dependent manner [11,12]. Our data suggest a common conformational gating mechanism by which the Ca²⁺-dependent inhibition of the α-class of gap junction proteins is mediated by the direct association of an intracellular loop region of these proteins with Ca²⁺–CaM [6,7,39,40].

Supplementary Material

supplement

Figure S1 Monitoring CaM–Cx50p^141–166 complex formation by MS

The MALDI–TOF-MS spectra of the free form of CaM (~17 kDa) and the complex form of CaM–Cx50p^141–166 (~20 kDa). The inset shows the molecular mass of Cx50p^141–166.

Figure S2 Monitoring the interaction between apo-CaM and Cx50p^141–166 by (¹H,¹⁵N)-HSQC spectroscopy

An overlay of HSQC spectra of apo-CaM (red) with the spectrum of the apo-CaM–Cx50p^141–166 (green).

Figure S3 Dansyl fluorescence anisotropy

D-CaM (1 μM) was titrated with Cx50p^141–166 in 50 mM Tris/HCl (pH 7.5), 100 mM KCl and 5 mM CaCl₂. The fluorescence anisotropy were measured at λ_ex = 335 nm and λ_em = 495 nm with an integration time of 20 s. The inset shows the anisotropy change for D-CaM (free), and the complex of D-CaM–Cx50p^141–166 (bound).

Figure S4 Comparison of the α-family of Cxs

An overlay of the HSQC spectra of holo-CaM (yellow), holo-CaM–Cx50p^141–166 (green) and holo-CaM–Cx43p^136–158 (violet). Arrows indicate signals that shifted to the same direction because of identical binding modes, whereas rectangles highlight some exceptions.

Abbreviations used

BAPTA-5N

5-nitro-1,2-bis-(o-aminophenoxy)ethane-N,N,N′,N′-tetra-acetic acid

[Ca²⁺]_i

intracellular Ca²⁺ concentration

[Ca²⁺]_o

external Ca²⁺ concentration

CaM

calmodulin

apo-CaM

Ca²⁺-chelated CaM

D-CaM

dansylated CaM

holo-CaM

Ca²⁺-saturated CaM

CaMBD

CaM-binding domain

CaMK

Ca²⁺/CaM-dependent protein kinase

CDZ

calmidazolium

CHCA

α-cyano-4-hydroxycinnamic acid

Cx

connexin

DTT

dithiothreitol

g_j

junctional conductance

HSQC

heteronuclear single-quantum coherence

MALDI

matrix-assisted laser-desorption/ionization

NTA

nitrilotriacetic acid

NTH

N-terminal helix

SA

sinapinic acid

TFE

trifluoroethanol

TM

transmembrane region

TOF

time-of-flight

V_j

transjunctional voltage