Novel Pharmacophores of Connexin43 Based on the “RXP” Series of Cx43-Binding Peptides
Gap junction pharmacology is a nascent field. Previous studies have identified molecules that enhance intercellular communication, and may offer potential for innovative antiarrhythmic therapy. However, their specific molecular target(s) and mechanism(s) of action remain unknown. Previously, we identified a 34-aa peptide (RXP-E) that binds the carboxyl terminal domain of Cx43 (Cx43CT) and prevents cardiac gap junction closure and action potential propagation block. These results supported the feasibility of a peptide-based pharmacology to Cx43, but the structure of the core active element in RXP-E, an essential step for pharmacological development, remained undefined. Here, we used a combination of molecular modeling, surface plasmon resonance, nuclear magnetic resonance and patch-clamp strategies to define, for the first time, a unique ensemble of pharmacophores that bind Cx43CT and prevent closure of Cx43 channels. Two particular molecules are best representatives of this family: a cyclized heptapeptide (called CyRP-71) and a linear octapeptide of sequence RRNYRRNY. These 2 small compounds offer the first structural platform for the design of Cx43-interacting gap junction openers. Moreover, the structure of these compounds offers an imprint of a region of Cx43CT that is fundamental to gap junction channel function.
Gap junctions are intercellular channels formed by oligomerization of connexin proteins. In the heart, the most abundant connexin is the 43-kDa isotype connexin (Cx)43. Cardiac gap junctions conduct electric impulses between cells to maintain normal rhythm, and their closure can be a substrate for cardiac arrhythmias.1 As such, drugs that selectively open gap junctions may offer a novel strategy for antiarrhythmic therapy and/or treatment of cardiovascular disorders.2–5
Gap junction pharmacology is a nascent field (see elsewhere for review6). Recently, hexapeptides such as AAP10 and its stable analog ZP123 (rotigaptide), together with a novel peptide, GAP134, have been found to modify gap junctional communication, and to show potential as antiarrhythmic agents.7–10 This accumulated evidence supports the notion of gap junction modification as a suitable pharmacological target.10 Yet, further development of these molecules is limited by the fact that their precise molecular target remains undefined, thus reducing their potential as platforms for target-specific drug design.11
As an alternative strategy, we have applied knowledge on the mechanisms of Cx43 chemical gating to design molecules that bind the carboxyl-terminal domain of Cx43 (Cx43CT) and modify its function. Gating of Cx43 relies on an intramolecular particle–receptor interaction between the C-terminal domain and the cytoplasmic loop.12–15 Using phage display, we identified a series of peptides containing the sequence “RXP” (arginine, any amino acid, proline) as a consensus Cx43CT binding motif and reported that a particular 34-aa peptide within this RXP series (dubbed RXP-E) binds to Cx43, prevents heptanol- and low pH–induced gap junction closure, and prevents action potential propagation block.16,17
Although these studies have shown significant and promising results, further applications of RXP-E are hampered because of the molecular size and low membrane permeability of this peptide, as well as the metabolic instability and poor oral bioavailability of peptides in general. Peptide mimetics, on the other hand, can be developed to retain the desired biological properties of a peptide. Steps in the design of mimetic molecules include identification of the essential active components (or amino acids) of the peptide sequence (the pharmacophores), determination of their structure/conformation in aqueous solution and finally, development of a corresponding pharmacophore model.18 Here, we have combined molecular modeling (based on structural analysis of the RXP series16) and experimental methods to identify the first group of pharmacophores (cyclized and linear peptides 6 to 8 aa long), that bind Cx43CT and prevent closure of Cx43 channels. This ensemble of pharmacophores represents a new platform for future development of small molecules with high efficacy and affinity that can prevent closure of gap junctions. Furthermore, we provide the first 3D imprint of a potential site in Cx43CT that can be used for binding of exogenous molecules.
Materials and Methods
Experimental methods for molecular modeling, electrophysiological experiments, surface plasmon resonance, and GST-binding assays, as well as nuclear magnetic resonance experiments followed standard, previously published procedures.16,19,20 Details (including statistical analysis) are provided in the Online Data Supplement at http://circres.ahajournals.org.
Cyclized Hexapeptides Based on Analysis of the RXP Series
Our previous studies revealed that 12-mer peptides that bind Cx43CT share at least 2 common features (see the previous report,16 including table 1 and online figure I of that article): the (1) presence of an RXP motif and (2) a predominance of basic amino acids. Thus, for identification of a core active element with Cx43CT binding activity, we modeled the structure of a peptide that fulfilled the following criteria: (1) was capable of binding Cx43CT, (2) showed an abundance of RXP sites, (3) was strongly basic, and (4) its primary sequence suggested a high-ordered secondary structure. A peptide labeled “RXP-4” by Shibayama et al16 best fulfilled these criteria. Analysis of its sequence (RRPPYRVPPKLF) showed 4 different RXP motifs (one of them in C-to-N direction, another placed terminally), a balance of charge of +4 at physiological pH (7.4), and a combination of 2 proline–proline repeats likely to induce turns in the peptide, thus facilitating a helix-like conformation. We therefore used RXP-4 as a starting point to identify a novel Cx43CT-binding platform. Molecular modeling predicted that, in an α-helical conformation, the proline–proline repeats in RXP-4 would face opposite directions (Figure 1A). If extended, this α-helix yielded 2 horseshoe conformations (Figure 1B), both containing RXP motifs. Separating the 2 horseshoe-like sequences between tyrosine in position 5 and arginine in position 6 (red dotted line in Figure 1B) yielded 2 shorter peptides, both having terminal RXP motifs and a high content of basic residues (RRPPY and RVPPKLF; see Figure 1B). Given that RRP is a terminal RXP motif in the original RXP-4 peptide, the RRPPY sequence was chosen for further analysis.
In a pentapeptide of sequence RRPPY, amino acids R1 and Y5 are placed at the opposite terminal ends, and the horseshoe conformation is anticipated to be at equilibrium with a corresponding “open” linear conformation (see Figure 1C). Thus, to stabilize the horseshoe conformation and keep the correct distance between R1 and Y5, the peptide was backbone-cyclized using an asparagine (N6) as a linker amino acid between R1 and Y5. The cyclic peptide (cyclo-RRPPYN) is illustrated in Figure 1D (small numbers mark the position given to each amino acid). This peptide was dubbed “CyRP-61” because it was the first cyclized, RXP-derived hexapeptide in this new series (see Table). Molecular modeling predicted an excellent correlation of the position in space of side chains R1R2Y5 in CyRP-61 (green in Figure 1E) with the equivalent R1R2Y5 side chains of RXP-4 (red in Figure 1E). Similar results were obtained by the conservative substitution N-Q at position 6 (CyRP-62; see the Table). This design strategy, previously used to transform peptide AAP10 to a more potent tripeptide analog, was predicted to generate more stable, smaller peptides with a core active structure capable of binding Cx43CT. This prediction was tested by surface plasmon resonance (SPR) experiments, as described below.
CyRP-62 Binds to Cx43CT
SPR allows for assessment of ligand–analyte binding in real time and was used for characterization of the RXP series.16 Cx43CT was covalently bound to the matrix of a sensor chip. Figure 2A shows plots of angle of incidence of resonance (in “response units”) as a function of time. Various concentrations of CyRP-62 were introduced in the microfluidics system at the time indicated by the upward arrow. Despite the small size of the analyte (798 Da), which approached the detection limits of the SPR instrument (≈500 to 700 Da), a clear change in resonance was recorded, thus indicating direct binding of CyRP-62 to Cx43CT. Dissociation ensued on washout (downward arrow). The rapid time course of association and dissociation precluded us from direct calculation of the dissociation constants. Yet, the results show that CyRP-62 binds Cx43CT. In previous studies, we demonstrated that peptide RXP-E prevents chemical gating of Cx43.16 As a next step, we tested whether this property was preserved in the CyRP peptides.
Junctional conductance was measured in N2a cell pairs expressing Cx43. Patch pipettes were filled with an internal solution containing the test peptide. Time course and extent of octanol-induced uncoupling was compared to that observed in the absence of the peptide. Figure 2B shows average measurements of junctional conductance (Gj) recorded at various times after the onset of octanol superfusion. Red symbols and line correspond to data obtained when the cyclized hexapeptide CyRP-62 was dissolved in the internal pipette solution, and black symbols correspond to the average time course of octanol-induced uncoupling in control conditions. The data show that, in the presence of the cyclized peptide, the progression of octanol-induced uncoupling was delayed, and 100% of cells remained coupled after 10 minutes after of octanol superfusion (see also Online Figure I, A). The minimum Gj value recorded at the end of the 10-minute octanol exposure was significantly different from that obtained in the absence of the peptide (P<0.01).
The results in Figure 2B suggested that CyRP-62 contains an active core sequence capable of interfering with octanol-induced uncoupling. Other studies have demonstrated the importance of aromatic side chains in the preservation of pharmacophore activity.18 Consistent with this notion, substitution of Y5 with a larger aromatic residue (W) disrupted the activity of the peptide (Figure 2C; also see the Table; peptide CyRP-63). Indeed, only 1 of 6 pairs remained coupled at the end of octanol exposure (Online Figure I, B), and the average Gj value was not statistically different from 0 (P>0.05). The side chains of N and Q differ in having only one and 2 methylene groups, respectively. Consistent with this structural preservation, the activity of peptide CyRP-61 (N in position 6) was similar to that observed for CyRP-62 (Figure 2D; Online Figure I, C). These data show that CyRP peptides represent a new class of core active molecules that interfere with Cx43 chemical gating. Additional improvement was achieved by increasing the balance of charge via addition of an arginine residue in the cyclized sequence.
Increased Charge: Cyclized Heptapeptides
One of the characteristic features of RXP peptides was the high balance of positive charge in their sequence (see16). Thus, an alternative cyclized peptide (in this case, a heptapeptide) was generated by separating the RXP-4 sequence between amino acids R6 and V7 (blue line in Figure 3A), then forming a backbone-cyclic peptide using either asparagine or glutamine to complete the structure (see also the Table; peptides labeled CyRP-71 and CyRP-72, respectively). These peptides, although somewhat larger, had an increased electrostatic balance (+3 instead of +2). As shown in Figure 3B, an overlay of CyRP-71 with CyRP-61 reveals that both peptides share similar coordinates for the position of amino acids R1, R2, and Y5, whereas CyRP-71 presents an additional arginine (R6). This simple modification had a significant impact on the binding Cx43CT, as demonstrated by the SPR experiment shown in Figure 3C (compare data with those in Figure 2A; notice different scale in the ordinates). When normalized by molecular weight, the amplitude of the SPR response elicited by CyRP-62 (125 μmol/L) was 3.71 (in arbitrary units); in comparison, the response elicited by CyRP-71 was 9 times higher (33.92 arbitrary units; immediately consecutive experiments conducted on the same sensor chip). In fact, the molecular mass–normalized response from CyRP-71 binding was larger than that obtained from RXP-E on the same sensor chip (9.91 arbitrary units), although there was a noticeable difference in the time course of the reaction (data not shown). Finally, patch-clamp results showed that CyRP-71 also prevented octanol-induced uncoupling (Figure 3D). In fact, there was an improvement in the ability of this core active component to prevent uncoupling after insertion of the basic residue, further underlying the importance of the balance of charge in the activity of these compounds (see also Online Figure I, D). Overall, this is the first demonstration of small cyclized peptides that can interfere with the octanol-induced closure of Cx43 channels.
Structural Correlation Between RXP-E and CyRP-71
In previous studies, we reported that RXP-E prevents chemically induced uncoupling of Cx43 channels, and action potential propagation block.16,17 Here, we tested whether structural features thought to be of relevance in the cyclized peptides were also present in RXP-E. Figure 4 (top) shows the entire RXP-E sequence (in the one letter amino acid code), divided into 3 regions: the first RXP-containing domain (the amino side; colored red), the linker (green), and the second RXP-containing domain (the carboxyl side; colored blue). Figure 4A and 4C shows the overlay of peptide CyRP-71 (green) with either the carboxyl terminus (amino acids 29 to 34) or the amino terminus (amino acids 1 to 11) of RXP-E (red), respectively. Clearly, the spatial coordinates of the basic residues R1, R2, and R6 of the cyclic peptide match with those of residues K29, R31 and R34 of RXP-E. The similar dimensions, spatial distribution and electric charge of the core residues make it likely for the carboxyl terminal region of RXP-E and the cyclic peptides to share a common binding motif. The opposite conclusion can be drawn from the comparison of the cyclic peptide with the amino end (N-terminal) of RXP-E; in that case, the space occupied by R6 in the cyclized peptide would correspond to the location of 2 acidic residues (D2, D3) in RXPE. If binding between Cx43CT and the peptide is mediated, at least in part, by electrostatic forces, the presence of negative charges at a relevant position in space would prevent binding. Overall, these modeling results led us to predict that binding of RXP-E to Cx43CT occurs via the carboxyl end of RXP-E. The SPR results shown in Figure 4B and 4D were consistent with this hypothesis. Duplicates of either the carboxyl side, or the amino side of RXPE, separated by the linker region, were used (see sequences at the top of Figure 4B and 4D). As shown in Figure 4B, a significant change in the angle of incidence for resonance was detected when a peptide containing the carboxyl end of RXPE was presented to Cx43CT (upward and downward arrows represent the addition of the peptide and its washout, respectively). In contrast, no binding was detected for the peptide formed by the amino end of RXP-E and its linker (Figure 4D). Overall, these results show a convergence between modeling predictions and experimental results and suggest defined structural constraints for the binding of RXP-derived peptides to Cx43CT. This structural information led us to the prediction of a minimum Cx43CT binding motif.
Identification of a Minimally Active Motif: RRNY
Cyclized compounds show longer bioavailability given their limited degradation in the intracellular space. An alternative path toward pharmacological development is the use of synthetic scaffolds that mimic peptide structure. Toward that aim, we sought to minimize the active sequence of the cyclized compounds by deleting the proline residues in CyRP-61 (Figure 5A). Indeed, in an α-helix, the proline–proline component would retain R1, R2 and Y5 at the appropriate spatial coordinates (see also Figure 1C). We therefore speculated that, because this spatial conformation was achieved by introduction of N6, the prolines would no longer be essential to hold the relative distances. Molecular modeling confirmed this expectation, as shown by the overlay of the side chains for R1, R2, and Y4 in a short tetrapeptide RRNY (green) with R1, R2, and Y5 of the cyclized CyRP-61 hexa-peptide (red; Figure 5B). Functional and biochemical assays were therefore designed to assess the prediction that RRNY is a Cx43CT binding motif. As a first step, we tested whether RRNY-containing peptides could bind Cx43CT. Assessment of binding was carried out by SPR. Recombinant Cx43CT was used as ligand, covalently linked to the sensor chip, and peptides were presented as analytes. The mass of the analyte was increased by concatenating 2 “RRNY” motifs (ie, RRNYRRNY). An example of sensograms obtained on introduction of this peptide into the chip containing Cx43CT, is presented in Figure 5C. Clearly, angle of incidence for resonance changed rapidly on onset of superfusion (time “zero”) and returned to baseline on washout (see downward arrow). The rapid transitions signaled a fast on–off ligand–analyte interaction and prevented us from determining dissociation constants for these peptides. However, the data showed that RRNYRRNY was capable of interacting with Cx43CT in a concentration-dependent manner. Similar results were obtained in 3 separate runs. Additional studies showed that RRNY-containing peptides can pull down Cx43 from an adult rat heart lysate preparation (Online Figure II).
The binding results led us to assess whether the motif RRNY would be sufficient for preventing octanol-induced uncoupling. Cx43-expressing N2a cells were dialyzed with an internal pipette solution containing peptide RRNYRRNY. Time course of octanol-induced uncoupling is shown in Figure 5D. In the presence of the peptide, average Gj recorded 10 minutes after the onset of octanol was significantly different from that recorded in control at the same time point (P<0.05) and coincided with the preservation of electric coupling in 4 of the 6 cells studied (see also Online Figure III, A). These results indicate that RRNY is a minimum sequence capable of interfering with the chemical regulation of gap junctions by octanol. Moreover, because amino acids RRNY are contiguous in the cyclized CyRP-61 peptide (see Figure 1D and the Table), we speculated that a linearized (noncyclic) RRPPYN peptide would fail to affect uncoupling, given the loss of continuity of the RRNY motif. As expected, the time course and extent of octanol-induced uncoupling in the presence of the linear RRPPYN peptide was not different from control (Figure 5D; trace in green; also Online Figure III, B), thus supporting the notion that pharmacophore activity is related not only to the presence of specific amino acids, but to the preservation of their molecular conformation in space.
CyRP-71 and the Integrity of the Cx43CT Domain
Figures 1 through 5⇑⇑⇑⇑ describe the various steps taken to identify potential leading compounds. Additional experiments focused on CyRP-71, given its binding and functional efficacy, and its potential as a more biologically stable compound. We hypothesized that binding of CyRP-71 to Cx43CT is linked to the ability of the peptide to prevent uncoupling. As shown in Figure 6A, this hypothesis was supported by experiments demonstrating that CyRP-71 failed to prevent octanol-induced uncoupling in Cx43 channels lacking the C-terminal domain (mutant M25716). In addition, this peptide did not modify the time course of octanol-induced uncoupling in N2a cells expressing a different connexin isoform, Cx40 (Figure 6B). Consistent with this observation, CyRP-71 caused only a minor SPR deflection when interacting with recombinant Cx40CT (Online Figure IV). The overall data suggest that there is a degree of structural specificity to the effect of CyRP-71 and that the Cx43CT domain is an essential component for peptide action.
CyRP-71 and Acidification-Induced Uncoupling
Octanol served as a screening tool to identify peptide activity. However, a more biologically relevant question is whether, as in the case of RXP-E, the candidate compound can prevent low pH-induced block.16 Experiments were conducted both in Cx43-expressing N2a cells (Figure 6C) and in pairs of neonatal cardiac myocytes (Figure 6D). Patch pipettes were filled with an internal solution buffered to pH 6.2. In the absence of peptide (black symbols and traces) Gj decreased progressively, reaching minima of 17.3±1.7% in N2a cells (Figure 6C) and 8.4±3.2% (Figure 6D) in cardiac myocytes. In contrast, cells exposed to CyRP-71 remained coupled throughout the same time course, Gj decreasing only to 49.0±5.9% (6C) and 47.0±9.5% (6D) of control (red symbols). The green symbols in Figure 6D depict data obtained with a linearized peptide RRPPYR. Results were not different from control, indicating that the structure of CyRP-71, rather than only the net balance of charge, was important for its functional effect.
CyRP-71 and Cx43CT Interaction Resolved by NMR
The SPR results demonstrated direct interaction between CyRP-71 and Cx43CT. Yet, further refinement of CyRP-71 will require a better understanding of the structural constraints imposed by its binding site in Cx43CT. As an initial step, we identified the amino acids in Cx43CT whose position in space change when presented to the peptide. Resonance assignments for Cx43CT have been published before (X). Figure 7A and 7B shows an 15N-HSQC spectrum for Cx43CT alone (black) which has been overlaid with spectra obtained in the presence of CyRP71 (red). A summary of the Cx43CT residues affected by CyRP-71 is presented in Figure 7C. The predominant regions of resonance shifts were: (1) near or within the proline-rich region of Cx43CT (K258-A276); (2) in the extreme end of the C-terminal domain (I382); and (3) in the region contained within amino acids G285 to E316 of Cx43. These results confirm the binding of Cx43CT to CyRP-71 and provide a first lead as to the amino acids in the C-terminal domain that are involved in the interaction. Future studies will be needed to determine which residues form the actual binding site and which shifts may result from secondary changes, perhaps some consequent changes in Cx43CT dimerization.21
We have previously demonstrated that RXP-E binds to Cx43CT and prevents chemically induced gap junction closure.16 As such, RXP-E represented a proof of principle, and the RXP series a starting point to develop a pharmacophore model for new compounds capable of modifying gap junctions. Structure–function analysis of the RXP series led us to a new ensemble of pharmacophores that bind Cx43CT and prevent chemically induced uncoupling. To our knowledge, the 2 core active structures hereby reported (one cyclized; one linearized) represent the smallest known Cx43CT-binding molecules with regulatory activity over gap junctions. Moreover, CyRPs are the first known cyclized molecules capable of binding Cx43CT. Thus, our results represent a potential foundation for development of target-based gap junction pharmacology.
Our compounds represent a new a pharmacophore model for Cx43 regulation. Yet, major milestones need to be reached before considering them of practical application. First, our studies demonstrate that selected peptides prevent closure of Cx43 channels induced by octanol superfusion. CyRP-71 also prevented pH-induced uncoupling. Yet, other assays will be needed to obtain a wider view of the overall functional effects of this peptide on Cx43. Second, we have shown that our candidate compounds interact with Cx43, but a remaining question is whether these peptides bind other cellular components (including other ion channel proteins) and alter their function and/or regulation. Additional studies will be required to characterize the selectivity and specificity of these compounds and their range of pharmacological applications. Yet, it is important to keep in mind that RXP-E, a leading molecule from the RXP series, prevents action potential propagation block without modifying other nonjunctional membrane channels,17 and our studies show structural similarities between the pharmacophores and the active domain of RXP-E. A lack of increase of Gj during application of both RXPE17 and CyRP71 may also indicate that these pharmacophores act not by recruiting new channels but by stabilizing the open state of those at the membrane. Third, we chose to explore the effect of CyRP-71 at one peptide concentration. Thus, we know that 100 μmol/L CyRP-71 delivered through the patch pipette is enough to prevent octanol, and low pH–induced uncoupling (Figure 6). Yet, further structure-based refinement of the pharmacophore will be needed to improve its affinity for the target. Fourth, the linkage between Cx43CT binding in SPR and the patch-clamp data may still seem correlative, although the hypothesis of a causative association is supported by results obtained with the mutant lacking Cx43CT (Figure 6A). These caveats notwithstanding, an important goal has been achieved: to find a compound of low molecular mass (<1 kDa) that binds the regulatory domain of Cx43, affects function, and shows selectivity for Cx43 over Cx40. Improving the affinity and selectivity of these compounds will make them valuable as potential scaffolds to act as carriers for cargoes with biological activity.
Our functional studies are complemented with NMR analysis of Cx43CT in the presence of CyRP-71. The data provide us with initial identification of those amino acids whose position in space is affected by the peptide (Figure 7). Differences were found with the resonance shifts caused by RXP-E.16 Of note, resonance shifts detected by HSQC may be consequent to direct binding, distant changes in conformation, and/or modifications of the dimerized state of Cx43CT.21 The larger, more complex structure of RXP-E (several “RXP” domains; likely more than one binding site) may cause a wider range of modifications, direct and indirect, that overlap (and may obscure) those occurring within the binding pocket. The cyclized peptide, on the other hand, may be less restricted for secondary interactions and therefore yield a different resonance shift map. Future identification of the actual binding pocket will be carried out by nuclear Overhauser effect spectroscopy (NOESY). This analysis will be an important step toward optimizing pharmacophore binding and selectivity. Overall, these peptides represent a potential imprint of a region of Cx43CT that is amenable for binding to exogenous molecules that will affect the function of the Cx43 channel as a whole. Combined structural and biological studies may lead to a future generation of molecules of higher efficacy, selectivity and biological stability, capable of crossing the cell membrane barrier to reach their target in a living cell. Although the latter is only a goal, our data suggest that we are heading in the right direction. Identification of the CyRP group of compounds bears relevance, because cyclic peptides are more stable within the intracellular space and, as such, have more potential for future pharmacological applications.
Peptide RRNYRRNY failed to prevent uncoupling in ≈40% of the cell pairs studied. From that standpoint, its efficacy was less than that previously described for RXP-E. On the other hand, CyRP-71 showed an effect similar to that of RXP-E. Interestingly, CyRP-71 also shows homology with the Cx43CT-binding element of RXP-E. This cyclized molecule offers itself as an excellent platform for the next generation of compounds, using peptide–mimetic substitutions on the core structure to minimize the size and maximize the activity, stability, and bioavailability while preserving pharmacological effect.
In summary, we have described a new generation of Cx43-binding peptides. Our efforts focused on a series of in silico modeling steps, combined with biochemical and cellular experiments, to identify the core active structure of the RXP series. We have identified new candidate molecules capable of binding Cx43CT and preventing chemically induced uncoupling of Cx43 channels. This is the first demonstration of a small, cyclic core active structure that chemically and functionally interacts with Cx43 to prevent gap junction closure. Our data opens a new line of investigation for development of target-based gap junction pharmacology.
Sources of Funding
Supported by NIH grants HL39707, HL087226, GM57691, and GM072631 (to P.L.S.) and a grant from the Carol M. Baldwin Research Fund (to S.M.T.).
B.D.L. holds modest ownership interest in the compounds detailed in this article, which were provided by Zealand Pharma.
Original received December 17, 2008; resubmission received May 8, 2009; revised resubmission received June 9, 2009; accepted June 12, 2009.
Srinivas M, Duffy HS, Delmar M, Spray DC. Prospects for pharmacological targeting of gap junction channels. In: Zipes DP, Jalife J, eds. Cardiac Electrophysiology: From Cell to Bedside. 4th ed. Philadelphia, Pa: Saunders; 2004: 158–167.
Eloff BC, Gilat E, Wan X, Rosenbaum DS. Pharmacological modulation of cardiac gap junctions to enhance cardiac conduction: evidence supporting a novel target for antiarrhythmic therapy. Circulation. 2003; 108: 3157–3163.
Lewandowski R, Petersen JS, Delmar M. Connexins as potential targets for cardiovascular pharmacology. In: Zipes DP, and Jalife J, eds. Cardiac Electrophysiology: From Cell to Bedside. 5th ed. Philadelphia, Pa: Saunders. 2009; 205–213
Rossman EI, Liu K, Morgan GA, Swillo RE, Krueger JA, Butera J, Gruver M, Kantrowitz J, Feldman HS, Petersen JS, Haugan K, Gardell SJ, Hennan JK. The gap junction modifier, GAP-134, improves conduction and reduces atrial fibrillation/flutter in the canine sterile pericarditis model. J Pharmacol Exp Ther. 2009; 329: 1127–1133.
Axelsen LN, Haugan K, Stahlhut M, Kjølbye AL, Hennan JK, Holstein-Rathlou NH, Petersen JS, Nielsen MS. Increasing gap junctional coupling: a tool for dissecting the role of gap junctions. J Membr Biol. 2007; 1: 23–35.
Delmar M, Coombs W, Sorgen P, Duffy HS, Taffet SM. Structural basis for the chemical regulation of connexin43 channels. Cardiovasc Res. 2004 62: 268–275.
Shibayama J, Lewandowski R, Kieken F, Coombs W, Shah S, Sorgen PL, Taffet SM, Delmar M. Identification of a novel peptide that interferes with the chemical regulation of connexin 43. Circ Res. 2006; 98: 1365–1372.
Lewandowski R, Procida K, Vaidyanathan R, Coombs W, Jalife J, Nielsen MS, Taffet SM, Delmar M. RXP-E: a connexin43-binding peptide that prevents action potential propagation block. Circ Res. 2008; 103: 519–526.
Oxford EM, Musa H, Maass K, Coombs W, Taffet SM, Delmar M. Connexin43 remodeling caused by inhibition of plakophilin-2 expression in cardiac cells. Circ Res. 2007; 101: 703–711.
Lang BD, Delmar M, Coombs W. Surface plasmon resonance as a method to study the kinetics and amplitude of protein-protein binding. In: Dhein S, Mohr FW, Delmar M, eds. Practical Methods in Cardiovasc Res. Heidelberg, Germany: Springer; 2005: 936–947.