Contribution of Intracellular Calcium and pH in Ischemic Uncoupling of Cardiac Gap Junction Channels Formed of Connexins 43, 40, and 45: A Critical Function of C-Terminal Domain

Ischemia is known to inhibit gap junction (GJ) mediated intercellular communication. However the detail mechanisms of this inhibition are largely unknown. In the present study, we determined the vulnerability of different cardiac GJ channels formed of connexins (Cxs) 43, 40, and 45 to simulated ischemia, by creating oxygen glucose deprived (OGD) condition. 5 minutes of OGD decreased the junctional conductance (Gj) of Cx43, Cx40 and Cx45 by 53±3%, 64±1% and 85±2% respectively. Reduction of Gj was prevented completely by restricting the change of both intracellular calcium ([Ca2+]i) and pH (pHi) with potassium phosphate buffer. Clamping of either [Ca2+]i or pHi, through BAPTA (2 mM) or HEPES (80 mM) respectively, offered partial resistance to ischemic uncoupling. Anti-calmodulin antibody attenuated the uncoupling of Cx43 and Cx45 significantly but not of Cx40. Furthermore, OGD could reduce only 26±2% of Gj in C-terminus (CT) truncated Cx43 (Cx43-Δ257). Tethering CT of Cx43 to the CT-truncated Cx40 (Cx40-Δ249), and Cx45 (Cx45-Δ272) helped to resist OGD mediated uncoupling. Moreover, CT domain played a significant role in determining the junction current density and plaque diameter. Our results suggest; OGD mediated uncoupling of GJ channels is primarily due to elevated [Ca2+]i and acidic pHi, though the latter contributes more. Among Cx43, Cx40 and Cx45, Cx43 is the most resistant to OGD while Cx45 is the most sensitive one. CT of Cx43 has major necessary elements for OGD induced uncoupling and it can complement CT of Cx40 and Cx45.


Introduction
Gap junctions (GJs) participate in traffic of signalling molecules and propagation of electrical impulse between adjacent cells by forming intercellular channels. Six connexin (Cx) molecules assemble to form cell surface hemichannel, whereas GJs are formed by the docking of two hemichannels from the adjacent cells [1]. Till date, more than 20 Cxs have been reported in human [2]. Cxs have four transmembrane domains, two extracellular loops, one intracellular loop and cytoplasmic N and C-termini [1]. In mammalian heart, 3 major Cx isoforms i.e. Cx45, Cx43 and Cx40 are expressed. Expression of Cx43 predominates in all parts of the heart, whereas expression of Cx40 and Cx45 are compartmentalized [3,4]. Overlapping expression pattern of Cxs raised the possibility of the formation of homomeric, heteromeric, homotypic and heterotypic junctions in heart [5][6][7][8]. Un-apposed hemichannels on the cell surface can function in certain physiological conditions, whereas GJs are constitutively active. The length of Cterminus (CT) varies among different Cx isoforms. CT plays a crucial role in the assembly, degradation and function of GJs. CT of Cx43 harbours multiple phosphorylation sites for several protein kinases [9,10].
During ischemia, gap junctional communication is compromised which may cause cardiac arrhythmia [11][12][13]. Interestingly, hemichannels and GJs behave differently in response to metabolic inhibition or simulated ischemia. Metabolic inhibition induces the opening of hemichannels but reduces the gap junctional coupling [14,15]. Uncoupling of GJs, limit the spread of ischemic damage [16]. The association of Cx43 in both ischemia-reperfusion injury and ischemic preconditioning, a mechanism by which repeated sub-lethal ischemia protects tissue from severe ischemia have been widely studied [17,18]. Ischemia-induced dephosphorylation of Cx43 and its translocation to the intracellular pool plays a major contribution in the inhibition of junctional communication [12]. Ischemia lowers the intracellular pH (pH i ) and increase cytoplasmic calcium ([Ca 2+ ] i ). Acidic pH i and elevated [Ca 2+ ] i have been shown to disrupt cell-cell communication in many cells including cardiomyocytes [19,20]. The inhibitory effect of [Ca 2+ ] i rise on junctional communication is possibly mediated through calmodulin [21]. Effect of intracellular acidification varies among different Cxs. Among the 3 cardiac Cxs, Cx45 and Cx40 showed highest and lowest pH-sensitivity, with Cx43 being intermediate [22]. CT of many Cxs dictates the pH sensitivity. Removal of CT impairs the pH gating of Cx43 and Cx40, but it had no effect on the pH sensitivity of Cx45. Further, a chimera of Cx40 having CT domain from Cx43 restored the pH sensitivity and vice versa [22].
Most of the studies relating to the ischemic uncoupling included 'whole heart' or 'isolated cardiomyocytes' where Cx45 and Cx40 containing GJs do exist, apart from Cx43 [12,13,23]. Therefore, the observed cumulative effect does not throw light on the susceptibility of individual Cxs to ischemia. Also, the molecular mechanism underlying ischemia mediated uncoupling of GJs is poorly understood. To study the effect of intracellular acidification on GJs, 100% CO 2 have often been used, which hardly mimic ischemic condition. Most studies, related to the effect of pH i or [Ca 2+ ] i on GJs studied individually in a non-ischemic milieu.
In the present study, we mimicked ischemia by creating oxygen glucose deprivation (OGD) and studied its effect on homomeric, homotypic GJ channels constituting of Cx45, Cx43 and Cx40, expressed in Neuro-2a (N2a) cells. The cumulative effect as well as individual contribution of pH i drop and [Ca 2+ ] i rise on the uncoupling of different GJs were studied in OGD condition. We observed that Cx45 is more susceptible to OGD, compared to Cx43 and Cx40. OGD caused maximum inhibition of junctional conductance (G j ) in Cx45 followed by Cx40 and Cx43. The role of CT in ischemic uncoupling was also studied by generating CT truncated mutants and swapping the CT between different Cxs.

Molecular biology
cDNA of mouse Cx45 and Cx40 was kindly provided by Prof. Klaus Willecke, Germany. Mouse eGFP tagged Cx43 (Cx43-eGFP) was provide by Prof. Feliksas F. Bukauskas, Albert Einstein College of Medicine, New York. Cx43, Cx45 and Cx40 were subcloned into pIRES2-DsRed or pIRES2-eGFP expression vector (Clontech, USA) by using standard molecular biology techniques. Cx45 was also sub-cloned into peGFPN1 vector (Clontech, USA) to express it as a fusion protein with eGFP. All the constructs were sequenced, followed by functional studies using patch clamp.
CT truncated Cx45, Cx43 and Cx40 were generated with the help of PCR based site directed mutagenesis by placing a stop codon at amino acid positions 272, 257 and 249 respectively, to generate Cx45-D272, Cx43-D257 and Cx40-D249. The primers used for generating CT truncated mutants are listed in table S1.
For creating CT-chimeric connexins, two step PCR based site directed mutagenesis was used according to previously published protocol [24]. Briefly, in the first PCR reaction, the Cx's CTdomain that has to be ligated with other Cx, was used as a c-DNA template. Primers having complementary sequences of both Cxs were used in the first PCR reaction. The amplified products of first PCR reaction bearing the complete CT part, was purified with the help of gel extraction kit (Qiagen, Germany) and used as megaprimers for the second PCR cycle. The Cx that contributes the Nterminus (NT) was used as template in the second PCR cycle to generate the CT-chimera. Constructs were confirmed by sequencing. The primer sets used for generating different chimeras are listed in table S2. Cx43-C40 represents chimeric Cx43 with the CT of Cx40; other chimeras are designated similarly.

Cell culture and transfection
N2a cells were obtained from National Centre for Cell Science, India. For routine maintenance, cells were grown in Dulbecco's Modified Eagle's Medium (DMEM), supplemented with 10% heat inactivated foetal bovine serum and 100 units/ml antimycotic and antibiotic mixture (Gibco, USA). Cells were maintained at 37uC with 5% CO 2 in a humidified incubator (Thermo Scientific, USA). For patch clamp recording, cells were grown on size-0 glass cover slips (Himedia Labs, India). Cells were transfected with desired Cx c-DNA constructs using Lipofectamine-2000 (Invitrogen, USA) in serum free media. After 4-5 hours of transfection, the serum free media was replaced with normal growth media containing 10% serum. All the experiments were performed after 24-36 hours of transfection.

Oxygen glucose deprivation (OGD)
Ischemia was simulated by exposing the cells to OGD, as described in earlier reports [25][26][27]. Cells grown on cover slips, were perfused first with bicarbonate external solution (ES; pH 7.4) containing (in mM) 124 NaCl, 4 KCl, 26 NaHCO 3 , 1.5 NaH 2 PO 4 , 1.5 MgSO 4 , 10 D-Glucose, 2 CaCl 2 . The solution was continuously bubbled with 5% CO 2 and 95% air, at room temperature (22-24uC). OGD was created by replacing the ES with ischemic solution (IS). IS has the same composition as ES with the exception of glucose being replaced with sucrose. IS was degassed for 1 hour followed by continuous bubbling with mixed gas containing 5% CO 2 and 95% of Argon. To assure complete removal of dissolved O 2 , O 2 scavenger sodium dithionite (2 mM) was added. Complete removal of dissolved O 2 was confirmed by analysing the IS with a dissolved O 2 analyser (Mettler Toledo, USA). To study the effect of ischemia, cells were exposed to IS for 5 minutes.

Patch clamp recording
For electrophysiology, N2a cells on cover slips were transferred to a recording chamber (RC-26G, Warner Instruments, USA), mounted on the stage of an Olympus IX71 inverted microscope, attached with EMCCD camera (Andor Technology, UK). The chamber was continuously perfused with ES. For measuring junctional conductance, cell pairs were patched with two Axopatch 200B (Molecular devices, USA) amplifiers. Thin walled glass pipettes of resistances 3-5 MV were prepared using pipette puller P-97 (Sutter Instrument Company, USA). The pipette solution contained (in mM) 10 NaCl, 140 KCl, 1 MgCl 2 , 0.2 CaCl 2 , 3 Mg-ATP, 5 HEPES (pH 7.2), 2 EGTA, unless otherwise mentioned. After obtaining whole cell, both cells were held at 0 mV. Holding potentials of cell-1 and cell-2 are designated as V 1 and V 2 respectively. To create junctional voltage gradient (V j = V 1 2V 2 ), V 1 was stepped to different voltages levels, keeping the V 2 at 0 mV. The current recorded from cell-2, represents the junctional current (I j = 2I 2 ). The current traces were low pass filtered at 1 kHz and sampled at 10 kHz with the help of Digidata 1440 (Molecular devices, USA).
In OGD experiments, V j of615 mV was applied to cell-1 for 5 second, with 5 second recovery step at 0 mV in between the pulses. The I j , obtained from cell-2, was normalized to 15 mV of V j , at every 10 th second to calculate G j . The maximum conductance, G jmax , of each experiment was used to normalize G j . The mean6SEM of normalized G j generated from 5-7 independent experiments were plotted against time, to check the degree of uncoupling due to OGD treatment.
To determine the voltage sensitivity of different Cxs, step protocol with an increment of 20 mV was applied to cell-1 ranging from 2120 mV to +120 mV for 20 s. There was a pause of 30 s at 0 mV between two sweeps. I j s obtained in response to 10 mV pre pulse were used to normalize the respective current traces. The normalized steady state G j (G j(ss) ) was plotted with respect to V j . To calculate voltage dependency, G j(ss) 2V j plot was fitted with two-states Boltzmann's equation that assumes channel gating is a two steps process. The following form of Boltzmann's equation was used: G j(ss) = [(G jmax 2G jmin )/{1+exp[A(V j 2V 0 )]}]+G jmin , where G jmax is the normalized maximum conductance (equal to 1), G jmin is the minimum conductance obtained at higher V j s, V 0 is the voltage at which G j(ss) is half maximal, A is the slope factor of the curve that can be defined as zq/kT, where z is valance of charge q that acts as voltage sensor for the transition from open to closed state. k and T represent Boltzmann's constant and absolute temperature, respectively.
Relaxation kinetics of WT and CT truncated mutants were compared to check the effect of truncation on channel gating. I j traces (recorded at 2100 mV, V j ) were fitted with standard single or double exponential equation by using pCLAMP 10.

Calcium imaging
Intracellular calcium was measured ratio-metrically with Fura-2, as described before [28]. Briefly, adherent N2a cells on cover slips were incubated with 10 mM Fura-2-AM (Molecular Probes, USA) for 30 minutes at room temperature (22-24uC), in solution containing (in mM) 140 NaCl, 5 KCl, 1 MgCl 2 , 10 HEPES (pH 7.4), 10 D-glucose, 2 CaCl 2 . After 30 minutes of incubation, cells were washed in fura-2 free buffer for 30 minutes. Fura-2 loaded cells were illuminated with dual excitation wave lengths of 340 nm and 380 nm, while capturing the emission signal at 510 nm with the help of appropriate excitation and emission filters (Chroma Technology, USA). Fast excitation switching was done by Lambda DG4 wave length switching system (Sutter Instrument Company, USA). Images were acquired with an EMCCCD camera (Andor Technology, UK) attached with Olympus IX71 microscope, controlled through IQ software. Images were acquired at every 5 second interval for about 10 minutes duration. There was no detectable bleaching of the dye during entire time course of the experiment. F340/F380 ratio was calculated from background subtracted images. Inclusion of 2 mM BAPTA in pipette solution interfered with fura-2AM based calcium imaging. In presence of BAPTA, Fura-2 fluorescence signal became erratic. To avoid this, Rhod-2 was used to detect [Ca 2+ ] i when BAPTA was present in pipette solution. Rhod-2AM loading process was similar to Fura-2-AM loading.
Intracellular pH (pH i ) measurement pH i was measured with 2',7'-Bis-(2-Carboxyethyl)-5-(and-6)-Carboxyfluorescein dye (BCECF) as described earlier [29,30]. Briefly, cells were incubated with 10 mM of BCECF-AM (Molecular Probes, USA) for 10 minutes in the same buffer, used for fura-2-AM loading, at room temperature (22-24uC). Experiments were performed after 10 minutes of washing in BCECF free buffer. BCECF loaded cells were excited at 440 nm and 490 nm; emission was captured at 530 nm using appropriate filter set (Chroma technology, USA). The ratio of the emitted intensity (F490/F440) was calculated off line after subtracting the back ground. Calibration of the ratio values (F490/F440) into pH i was done by the equation: Where R is the ratio (F490/F440), R pH 7 is the ratio at pH 7, R min , R max and pK are the minimum ratio, maximum ratio and negative log of the dissociation constant respectively. For calibration, BCECF loaded cells were perfused with high K + solution of different pH ranging from 4 to 9, containing (in mM) 105 Potassium Aspirate, 4 KCl, 1.5 KH 2 PO 4 , 1.5 MgSO 4 , 49 NMDG, 10 EGTA, 10 glucose, 10 HEPES with 0.01 nigericin (Invitrogen, USA). The average (n = 35) ratio values (490 nm/ 440 nm) obtained from pH 4 to pH 9 were fitted with non-linear logistic equation to determine the values of R min , R max and pK. The value of R pH = 7 was obtained at the end of each experiment, by adding 10 mM nigericin to high K + solution of pH 7.
For clamping pH i at a particular value during OGD, pipette solution with increasing concentrations of HEPES (5, 10, 20, 50 & 80 mM) was used. With increasing concentration of HEPES, equimolecular amount of KCl was removed to maintain the osmolarity. The pipette solutions were usually supplemented with 50 mM BCECF free acids (Molecular Probes, USA) to avoid dilution of the BCECF dye inside the cell. At the onset of the experiments, the pipette solutions were adjusted to pH 7.2 with KOH.

Statistical analysis
Data provided here are representative of 3-5 independent experiments. Values are mean6SEM of 5-11 replicates as described in the respective figure legends. Student paired t-test and one way ANOVA were performed for comparison between two and multiple groups respectively.

Biophysical characterization of wild type and eGFP tagged gap junctions
Wild type Cx43, Cx40 and Cx45 readily formed GJs when expressed in N2a cells. Figure 1 shows the representative current traces at different junctional voltages (V j ), recorded from the cell pair containing homotypic junctions of Cx43, Cx40 and Cx45. The average junctional conductance (G j ) for Cx43, Cx40 and Cx45 were 6064 nS (n = 45), 4365 nS (n = 46) and 3664 nS (n = 48) respectively (Table 1). G j(ss) 2V j dependence of Cx43, Cx40 and Cx45 are presented in the lower panel of figure 1. A 10 mV pre-pulse was applied before every voltage step to normalize the G j(ss) . The normalized G j(ss) (with respect to 10 mV pre pulse) versus V j plots were fitted with two state Boltzmann equation to determine the voltage sensitivity. The Boltzmann parameters are presented in table 2. V 0 represents the V j at which G j(ss) is reduced by 50%. In agreement with previous reports, V 0 for Cx45 (+1161.8 and 21161.5) was lowest among all three Cxs (Cx43: +5760.5 and 25660.5, Cx40: +5060.7 and 24961.2), indicating its highest V j sensitivity. The V j sensitivity followed the order Cx45.Cx40.Cx43.
GJs are often studied by attaching different fluorescent proteins at the CT of Cx [31]. We analysed the V j sensitivity of Cx43-eGFP and Cx45-eGFP as these were used in our study. As shown in figure S1, G j(ss) 2V j plot of Cx45-eGFP is indistinguishable from the wild type Cx45. On the other hand, G j(ss) 2V j plot slightly shifted towards higher voltage upon attachment of eGFP with Cx43 indicating the decrease of voltage sensitivity, which is consistent with earlier reports [31]. V 0 (in mV) for Cx43 were +5760.5 and 25660.5, whereas for Cx43-eGFP the values were +6261.4 and 26261.2 respectively (Table 2). Cx40-eGFP was Figure 1. Sensitivity of Cx43, Cx40 and Cx45 to junctional Voltage (V j ). Upper panel shows the voltage step protocol. V j steps of 20 mV increment, from 2120 to +120 mV were applied for 20 sec followed by 30 sec recovery at 0 mV. A short pre pulse (100 msec) of 10 mV was applied at the beginning of each pulse to normalize G j . Ischemia uncoupled Cx43, Cx40 and Cx45 to different extents Ischemia was simulated by exposing the cells to OGD for 5 minutes. Cell pairs expressing Cx43, Cx40 or Cx45 were voltage clamped with two patch pipettes in whole cell configuration. Junctional current (I j ) from cell-2 was monitored continuously in external solution (ES), bubbled with 5% CO 2 +95% air, by stepping cell-1 from holding potential, 0 mV to615 mV (for 5 sec) in every 5 sec interval. Cell-2 was clamped at 0 mV. After confirming that there was no spontaneous run down of I j , ES was replaced with ischemic solution (IS), bubbled with 5% CO 2 +95% argon. After 5 minutes, IS was replaced back with ES. When cells were exposed to OGD, I j of all Cxs reduced immediately. I j reached a steady value within 5 minutes and did not decrease further under prolonged OGD treatment. Figure 2A shows the representative I j of Cx43, which decreased to almost 50% within 5 minutes of OGD. Interestingly the extent of uncoupling varied among Cxs. As shown in figure 2B and 2C, Cx43 was quite resistant to OGD in comparison to Cx45 and Cx40. Cx45 showed the highest sensitivity to OGD. In case of Cx45, G j decreased by 8562% (n = 9) in 5 minutes OGD, whereas there was only about 5363% (n = 7) reduction of G j for Cx43 and 6461% (n = 7) for Cx40. The rate of uncoupling also varied among Cxs. As shown in figure 2B, Cx43 uncoupled at a much slower rate in comparison to Cx45 and Cx40. Tagging of GFP to CT of some Cxs has been reported to alter their properties. For example, Cx45-eGFP, failed to rescue the Cx45 knockout mice from embryonic lethality [32]. We compared the susceptibility of Cx43-eGFP and Cx45-eGFP to OGD with their wild type counterpart. Effect of OGD on Cx43-eGFP was same as Cx43 (data not shown). However Cx45-eGFP showed higher sensitivity to OGD. As shown in figure 3, G j of Cx45-eGFP reduced by 9960.2% (n = 8) in 5 minutes of OGD treatment.
[Ca 2+ ] i rises and pH i decreases during simulated ischemia Intracellular acidification and [Ca 2+ ] i rise are known to be associated with ischemia [26]. Fura-2 loaded N2a cells when exposed to OGD, [Ca 2+ ] i increased immediately and maintained a steady value through the entire duration of OGD ( Figure 4A).
[Ca 2+ ] i returned to basal level when IS was replaced with ES. F340/F380 increased to 1.660.2 from 0.9260.1 (n = 25). Cx transfected cells also showed [Ca 2+ ] i rise to the same extent and there was no difference among Cxs (data not shown).
pH i was estimated ratio-metrically with BCECF. pH i dropped from 7.2560.1 to 6.3560.4 (n = 20) in 5 minutes of OGD. pH i was restored once IS was replaced with ES ( Figure 4B). The degree of intracellular acidification does not depend on the level Cx expression. The pH i of untransfected cells also dropped to the same extent following 5 minutes of OGD (data not shown).  Figure 5A) and attenuated the reduction of G j  of all Cxs significantly ( Figure 6). After 5 minutes of OGD, G j of BAPTA treated cell pair reduced by: 3562% (n = 6) for Cx43, 5262% (n = 6) for Cx40 and 5964% (n = 7) for Cx45. In control experiment (without BAPTA), G j of Cx43, Cx40 and Cx45 decreased by 5363%, 6461% and 8562% respectively. It is evident that [Ca 2+ ] i has lesser contribution in the uncoupling of Cx40 in comparison to Cx43 and Cx45. Effect of BAPTA is independent of pH i changes. It did not affect intracellular acidification during OGD ( Figure 5B).
To study the role of acidification on the uncoupling of GJ, we restricted the pH i change by using high concentration of HEPES in pipette solution. As shown in figure 5C, gradual increase of HEPES concentrations from 5 mM to 80 mM, prevented the pH idrop in a graded fashion. 80 mM of HEPES completely prevented acidification and pH i was maintained at 7.2 throughout the OGD. However it did not prevent the [Ca 2+ ] i rise, as represented in figure 5D. Like clamping of [Ca 2+ ] i , when pH i was maintained to 7.2, all Cxs showed lesser vulnerability to ischemic uncoupling. Reduction of G j was prevented significantly (Figure 7). pH iclamped cell pair showed the reduction of G j : 2661% (n = 7) for Cx43, 2962% (n = 6) for Cx40 and 2761% (n = 7) for Cx45.
To evaluate the cumulative role of [Ca 2+ ] i and pH i , we clamped both by injecting high concentration of potassium phosphate buffer. Addition of 150 mM KPO 4 in pipette solution prevented acidification as well as [Ca 2+ ] i rise during OGD, as shown in figure 5B and 5D. When [Ca 2+ ] i and pH i were clamped, OGD could not uncouple any of the junctions as observed by the steady G j throughout (Figure 7). Cx43, Cx40 and Cx45 became completely resistant to OGD.

Calmodulin (CaM) is involved in ischemic uncoupling
CaM has been reported to modulate both voltage and chemical gating of gap junctions [33,34]. CO 2 mediated uncoupling of Cx45 was prevented by suppressing the expression CaM with antisense RNA [35]. We studied the role of CaM in ischemic uncoupling. Anti-CaM antibody (Sigma-Aldrich, USA) was injected into the cell pair through patch pipette by supplementing pipette solution with 7 mM of antibody. Antibody was allowed to diffuse for 5 minutes after obtaining whole cell and I j was recorded. As shown in figure 6, the antibody attenuated ischemic uncoupling of Cx43 and Cx45, GJs significantly. However, uncoupling of Cx40 by OGD was not affected at all by anti-CaM antibody. G j of the antibody treated Cx43 expressing cell pair reduced by 2662% (n = 7) in response to 5 minutes OGD, whereas in control (without anti-CaM antibody) the reduction was 5363%. In case of Cx45, antibody treated and the control cell pair showed the reduction of G j by 5462% (n = 8), and 8562% respectively ( Figure 6D).

Role of carboxy terminal (CT) tail in ischemic uncoupling
Properties of CT deletion mutants. CT of Cxs has been implicated in many physiological and pathophysiological processes [22,32,36]. To study the role of CT in ischemic uncoupling, we generated CT-deletion mutants Cx43-D257, Cx40-D249 and Cx45-D272 (see material and methods). Deletion of CT impaired the G j of Cx40 and Cx45. The average G j of Cx40 reduced from 4364.6 nS (n = 46) to 460.5 nS (n = 20) upon deletion of CT. Similarly, Cx45-D272 showed average G j of 360.5 nS (n = 23) which is about 12 fold lesser than wild type (3663.5 nS, n = 48), (Table 1). Consistence with earlier reports, Cx43-D257 displayed higher average G j and larger plaque diameter than the wild type ( Table 1). Deletion of CT also altered the gating parameters of GJs. In accordance with a previous report, removal of CT from Cx43 and Cx40 did not alter their sensitivity to V j ( Table 2) as V 0 remained unaltered. However there was a small change of minimum conductance (G min ) and the desensitization kinetics altered significantly (Table 2). To study the desensitization kinetics, V j was stepped to -100 mV. The decay of I j was fitted with either single or double exponential function as required. The time constants for relaxation, t 1 and t 2 , are presented in figure S2 and table 2. Most GJs are known to follow two gating processes: slow gating and fast gating [31,37,38]. In accordance with previous reports [37,38], all wild type GJs in our experiment showed double exponential decay of I j (Figure 2 and Table 2). Fast time constant and slow time constant of voltage relaxation reflect closure of fast gate and slow gate respectively [37]. Cx43-D257 and Cx40-D249 followed monophasic decay compared to biphasic relaxation of the corresponding wild types, suggesting loss of one gate in both cases  [37,38]. Interestingly removal of CT did not affect the gating of Cx45 as both slow and fast t remained unaltered ( Figure S2 and Table 2).
Immunofluorescence analysis revealed the characteristics of junctional plaque formed by different CT-truncated mutants (Figure 8). Both wild type and Cx43-D257 showed bright fluorescent plaque at the cell-cell junction. Average diameter of the Cx43-D257 junctional plaque is significantly bigger than the wild type Cx43 (Table 1), which is in agreement with the observed higher G j of Cx43-D257. Although clear junctional plaques were detected in cell pairs containing wild type Cx40 or Cx45, it was not visible in corresponding CT truncated mutants ( Figure 8); a possible reason for their very low G j .

Effect of OGD on CT truncated mutants
Cx43-D257 when exposed to OGD, showed significant resistance to uncoupling. 5 minutes of OGD resulted in only 2662% (n = 5) reduction of G j , compared to 5363% in wild type (Figure 9), suggesting the involvement of Cx43-CT in ischemic uncoupling. OGD on deletion mutants of Cx40 and Cx45 could not be performed due to their low average G j .
We observed that among 3 cardiac Cxs, Cx45 is the most sensitive to OGD, whereas Cx43 is the least one. To check if this variability is conferred by the CT, we generated several Cx chimeras e.g. Cx43-C45, Cx43-C40, Cx40-C45, Cx40-C43, Cx45-C43 and Cx45-C40 by swapping CT. Biophysical properties of different chimeras and their sensitivity to OGD are described below.

Biophysical properties of CT chimeras
Deletion of the CT of Cx40 and Cx45 attenuated their G j by more than 10 fold (Table 1). CT truncated mutants Cx40-D249 and Cx45-D272 did not assemble properly to form junctional plaque as evident from confocal microscopy. When CT of Cx43 was tethered to truncated Cx40 and Cx45, it rescued their G j completely (4063.6 nS, n = 35 for Cx40-C43 and 4964.6 nS, n = 37 for Cx45-C43), (Table 1). Cx45-C43 even showed higher G j than wild type Cx45 (3663.5, n = 48). Also, both the chimeras showed bright junctional plaques with diameters comparable to  Table 1). Tethering of CT of Cx40 to the Cx43-D257 restored its G j to the wild type value ( Table 1). CT of Cx40 also rescued Cx45-D272. Cx45-C40 chimera behaved like wild type Cx45 in terms of G j (3764.3 nS, n = 33 for Cx45-C40 and 3663.5 nS, n = 48 for Cx45), though the plaque diameter was slightly smaller (Figure 8 and Table1). Interestingly, CT of Cx45 was not compatible with either Cx43 or Cx40. None of the chimeras with CT of Cx45 formed functional channels ( Figure 8 and Table 1).
The voltage sensitivity and relaxation kinetics of different CT chimeras was studied by step protocol. The results are shown in table 2. Except Cx45-C40, voltage sensitivity of all chimeras resembled their respective wild type. V 0 of Cx45-C40 shifted towards higher voltage indicating decrease of V j sensitivity. Like wild type Cx, voltage relaxation of all chimeras showed biphasic decay of I j . In comparison to wild type, all chimeras except Cx45-C40 showed slower decay as indicated by higher t 1 and t 2 . t 1 and t 2 of Cx45-C43 are not significantly different from wild type Cx45.

Ischemic uncoupling of CT chimeras
Effect of 5 minutes OGD on different chimeras, are presented in figure 10. If the CT of a particular Cx is the sensor for ischemia, chimera would respond like the Cx that contributed the CT. Cx43-D257 showed 2662% inhibition of G j in response to OGD. Tethering of CT of Cx40 did not improve the sensitivity to OGD significantly. G j of Cx43-C40 inhibited by 3362% (n = 7), whereas wild type Cx40 exhibited 6461% inhibitions. Similarly, CT of Cx40 though rescued G j of truncated Cx45; the chimera exhibited far lesser sensitivity to OGD than Cx40 and wild type Cx45. G j of Cx45-C40 reduced by 3861% (n = 7) as shown in figure 10. Interestingly, CT of Cx43 not only rescued the G j of truncated Cx45 and Cx40, it brought their OGD sensitivity closer to wild type Cx43. After 5 minutes of OGD, Cx43, Cx40-C43 and Cx45-C43 showed reduction of G j by 5363% (n = 7), 4462% (n = 8) and 4561% (n = 6) respectively. It suggests that unlike CT of Cx40, CT of Cx43 is partially associated with OGD mediated uncoupling.

Discussion
Ischemia reduces gap junctional communication in many cell types. Intracellular acidification and [Ca 2+ ] i rise, generally observed during ischemia, are known to promote uncoupling of GJs. However the molecular mechanisms are not clearly understood. In the present study, we simulated ischemia by depriving oxygen and glucose and studied its effect on cardiac GJ channels, formed of Cx43, Cx40 and Cx45. Contribution of [Ca 2+ ] i and pH i in uncoupling was studied. Additionally we demonstrated the role of CT of Cxs in OGD mediated uncoupling.
We studied the effect of ischemia on different cardiac gap junctions, over-expressed in N2a cells. N2a cells have been used extensively in many earlier reports for studying different connexins as they do not have endogenous gap junctions [39,40]. Also, N2a cells are good model system for studying ischemia [41,42]. Cardiomyocytes or cell line of cardiac origin would have been an appropriate model for studying cardiac connexins. However, cardiomyocytes express all three connexins (Cx45, Cx43 and Cx40) which may form homomeric and heteromeric channels in different combinations, thereby making it difficult to study the effect of ischemia on individual connexins. Moreover, available gap junction blockers are not very specific. Similarly, all available cell lines of cardiac origin express multiple connexins.
GJs made of Cx45, Cx43 and Cx40 varied remarkably in their biophysical properties as well as sensitivity to OGD. In accordance with earlier report [43], we showed V j sensitivity follows the order Cx45.Cx40.Cx43. V j sensitivity is an inherent property of GJs and determined by several factors including number of charge residues in the voltage sensor. Voltage sensors in cardiac GJs are ill defined and the underlying mechanism of differential voltage sensitivity is unknown. Voltage sensitivity of Cxs may play an important role during ischemic uncoupling. Ischemia affected cardiomyocytes, depolarize quickly which may cause development of higher V j between ischemic cell and healthy cell. GJs that are less sensitive to V j , will remain open longer and may prevent development of arrhythmia.
OGD reduced gap junctional communication of all cardiac Cxs. However, the degree of uncoupling varied among Cxs. None of the GJs closed completely upon exposure to OGD for 5 minutes. Cx45 showed maximum reduction of G j , while Cx43 was most resistant to OGD. Once cells were exposed to IS, there was elevation of [Ca 2+ ] i and pH i dropped. When [Ca 2+ ] i rise was prevented with BAPTA without altering the acidification, all GJs showed partial resistance to ischemic uncoupling. In calcium clamped condition, when pH i was allowed to drop to the same extent, G j of Cx40 reduced by 5262% which is close to control value, suggesting that acidification is the major contributor in uncoupling. In the same condition, G j of Cx43 and Cx45 reduced by 3562% and 5964% compared to 5363% and 8562%  were significantly bigger (p,0.05) than that of wild type Cx43. There was no significant difference in plaque diameter between Cx43-C40 and Cx43. Cx43-C45 did not form detectable plaque. Most of the connexins were detected in the perinuclear area (shown with white arrows). B. I, II, III and IV are the images of Cx40, Cx40-D249, Cx40-C43 and Cx40-C45 expressing cells respectively. Cx40-D249 plaques were not detectable. Plaques formed by Cx40-C43 were thin and punctuated (shown with white arrows). Cx40-C45 did not form detectable plaque. C. I, II, III and IV, images of Cx45, Cx45-D272, Cx45-C43 and Cx45-C40 transfected N2a cells respectively. Cx45-D272 junctional plaques were not detectable. Cx45-C43 plaques were significantly larger than corresponding Cx45 plaques. Cx45-C40 plaques were punctuated and smaller compared to wild type Cx45. doi:10.1371/journal.pone.0060506.g008 reduction in control ( Figure 6 and Table 3). When pH i change was restricted without affecting [Ca 2+ ] i rise, all Cxs showed only 26-29% reduction of G j suggesting that acidosis indeed contributes more than [Ca 2+ ] i elevation in uncoupling. When both pH i and [Ca 2+ ] i were maintained to the normal physiological levels, OGD had no effect on any of the GJs (Figure 7 and Table 3). It confirms that acidosis and [Ca 2+ ] i elevation are the primary causes of ischemic uncoupling of GJs. There are ample evidences suggesting that [Ca 2+ ] i works through CaM [21]. To check the involvement of CaM in ischemic uncoupling of GJ, we administrated anti-CaM antibody intra-cellularly through patch pipette. In case of Cx43, anti-caM antibody prevented uncoupling to a greater extent than BAPTA, suggesting its additional role, independent of calcium ( Figure 6 and Table 3). CaM has been reported to bind to lens gap junction protein in calcium independent fashion [44]. In case of Cx45, buffering of [Ca 2+ ] i and inactivating CaM prevented the uncoupling to the similar extent, suggesting that [Ca 2+ ] i primarily worked through CaM in this case. Anti-CaM antibody had no effect on the uncoupling of Cx40. Unlike Cx43 and Cx45, Cx40 does not have putative CaM binding domain [45]. Therefore CaM promotes ischemic uncoupling of Cx43 and Cx45 but not in Cx40. It is not clear how CaM binding induces uncoupling [21,45]. It has been proposed that CaM occludes channel mouth of Cx32 from cytosolic side inducing channel closure [20,33]. This may be true also for Cx43 and Cx45. From the above discussion, it is apparent that lesser sensitivity to acidic pH makes Cx43 more resistant to OGD, compared to Cx45 and Cx40.
To investigate the role of CT in ischemic uncoupling, we generated different CT truncated Cxs and chimeras. CT deletion mutant Cx43-D257 when exposed to OGD, uncoupling was attenuated significantly, suggesting its involvement in uncoupling ( Figure 9). Earlier studies demonstrated that the intracellular loop, but not the CT of Cx43, is involved in calcium-CaM mediated reduction of G j [45,46]. Therefore, observed uncoupling in Cx43-D257 is possibly executed through calcium-CaM pathway. In the same line, OGD reduced the G j of Cx43-D257 to the same extent of wild type Cx43, when pH i was clamped in the latter. This suggests that CT of Cx43 is involved in acidosis mediated component of the uncoupling. This is consistent with the findings that CT-truncated Cx43 showed lesser sensitivity towards low pH, and resulted in increase of infarct size and arrhythmia due to acute coronary occlusion [36]. Upon tethering of CT of Cx43 to the truncated Cx40 and Cx45, not only did it improve their expression (Figure 8), but also the chimeras showed moderate sensitivity to ischemia ( Figure 10). It was not possible to assess directly the contribution of CT of Cx40 and Cx45 in uncoupling, due to the low expression of truncated mutants. When CT of Cx40 was tethered to Cx43-D257, the sensitivity of the chimera to OGD did not improve much, suggesting its minor role in ischemic uncoupling ( Figure 10). Cx45 chimera containing CT of Cx40 showed remarkably improved expression ( Figure 8). However, OGD reduced the G j of Cx45-C40 only by 3861%, compared to 8562% for Cx45 and 6461% for Cx40. OGD mediated uncoupling of Cx40 is mainly due to acidosis and calcium-CaM had minimum contribution in it. If CT of Cx40 is involved in the reduction of G j , more robust uncoupling of Cx45-C40 would have been observed. Therefore CT of Cx40 possibly does not participate in ischemic uncoupling. We could not assess the role of CT of Cx45 in ischemic uncoupling as tethering of it with Cx43 or Cx40 did not yield a functional construct.
Apart from their role in ischemic uncoupling, we observed that the CT determines the characteristics of junctional plaque ( Figure 8). In concurrence with previous report [47], removal of CT increased the plaque diameter made of Cx43. The G j also improved ( Table 1). However we could not detect junctional plaque in case of CT truncated Cx45 and Cx40. They also exhibited very little G j , possibly due to removal of phosphorylation and protein-protein interaction sites. Attachment of CT of Cx43 to truncated Cx40 and Cx45 enabled them to form junctional plaque with some altered properties. Plaque diameter of Cx40-C43 was smaller than Cx40 and appeared punctuated but the G j of both were comparable. Intriguingly, tethering CT of Cx43 to truncated Cx45 resulted in bigger plaque than wild type Cx45 and increased G j . Similarly Cx45, when tethered to CT of Cx40, it formed distinct plaque and functional GJ. It suggests that CT of Cx43 is an essential component required for the expression and formation of junctional plaque and it can complement CT of Cx40 and Cx45. However it is not required for the formation of its own junctional plaque. CT of Cx40 is necessary for its expression and plaque formation; it can also complement CT of Cx45. In contrary, CT of Cx45 does not complement Cx40 or Cx43.
In accordance with previous reports, CT domain showed its involvement in voltage gating [37,38,42]. Truncation of CT reduced G min and changed slope factor in case of Cx43 and Cx40. Following voltage step, I j of CT-truncated Cx43 and Cx40 decayed mono-exponentially compared to biphasic decay of the corresponding wild type. Relaxation kinetics became slower for Cx43-D257, while it accelerated for Cx40-D249, as reported earlier [37,38]. But truncation of the CT of Cx45 did not affect voltage dependent gating or relaxation kinetics ( Figure S2). The only change observed was reduced surface expression and junctional plaque formation, resulting in decreased G j .

Conclusions
This work figures out the molecular mechanisms that uncouple GJ channels formed of Cx43, Cx40 and Cx45 during ischemic condition. Ischemia was simulated by depriving oxygen and glucose. OGD caused the reduction of G j of all Cxs significantly. Cx45 was most sensitive, while Cx43 showed maximum resistant to OGD. Elevated [Ca 2+ ] i and acidic pH i were the primary causes, as uncoupling was prevented completely by restricting their change. Acidosis has greater contribution than elevated calcium in reducing the G j , particularly in the uncoupling of Cx40. Calcium worked through CaM, though calcium independent role of CaM in the uncoupling of Cx43 cannot be ruled out. Further, CT of Cx43 played significant role in ischemic uncoupling. Tethering of it to the CT-truncated Cx45 and Cx40 enabled them to respond to OGD, same like Cx43. CT also played significant role in determining plaque diameter, voltage dependent gating and current relaxation kinetics with the exception of Cx45-CT domain. Taken together, this study provides an explanation for the comprehensive mechanism of the ischemic uncoupling of cardiac gap junctions.

Supporting Information
Figure S1 V j sensitivity of wild type and eGFP tagged Cx43 and Cx45. A. Tagging of eGFP to the C-terminus of Cx43 decreases its voltage sensitivity. The G j(ss) -V j plot shifted towards higher voltage. B. Voltage sensitivity of Cx45 did not change after tagging eGFP. G j(ss) -V j plot of Cx45-eGFP is indistinguishable from that of wild type Cx45. (TIF) Figure S2 Desensitization kinetics of wild type and CTtruncated connexin containing gap junctions. AI and AII, representative I j traces of Cx43 and Cx43-D257. V j was stepped to 100 mV. I j decayed mono-exponentially and bi-exponentially for truncated Cx43 and wild type Cx43 respectively. Predicted best fittings are presented with solid line. AIII, merged fitted curves of