Heterocellular Contact at the Myoendothelial Junction Influences Gap Junction Organization
Heterocellular communication between vascular smooth muscle cells (VSMC) and endothelial cells (EC) at the myoendothelial junction (MEJ) is a critical part of control of the arteriolar wall. We have developed an in vitro model of the MEJ composed of primary cultures of murine EC and VSMC. Immunoctytochemistry and immunoblots demonstrated Cx37 and Cx43 in both cell types, whereas Cx40 was found only in EC. Cx37 was excluded from the MEJ in both EC and VSMC. Connexin composition as well as functionality of the gap junctions at the MEJ was assessed by measuring diffusional transfer of biocytin and Cy3. Using connexin-specific blockers and manipulations of expression of individual connexin proteins, we confirmed that Cx37 is not a part of EC–VSMC coupling, and we demonstrated that heterotypic gap junctions are functional at the MEJ. We speculate that specific gap junction organization may be a vital component of EC–VSMC contact at the MEJ.
Heterocellular communication, between and among cells of the vessel wall, plays a critical role in many aspects of vascular function and pathology.1,2 One form of communication between endothelial cells (EC) and vascular smooth muscle cells (VSMC) appears to be via extensions of the 2 cell types that pass through the internal elastic lamina and make contacts called myoendothelial junctions (MEJ).3–5 Gap junctions at the points of cell–cell contact in the MEJ form pathways for flow of second messengers or electrical current between the two cells,4,6–8 and this communication is critical in unifying their function.
Although the physiological importance of the MEJ is well known, research on this structure is limited because of the anatomical location of MEJ, limited access to both cell types in vivo, bioavailablity of inhibitors in vivo, and isolation of sufficient quantities of protein from both cell types.2 In view of these complexities, an in vitro model would be highly desirable. We, therefore, developed such a model from primary cultures of mouse EC and VSMC. We used immunoblots to identify the connexins expressed and immunocytochemistry to identify connexin proteins trafficked to the MEJ. In addition, we measured movement of diffusible solutes through the gap junctions at the MEJ and used a combination of gap junctional blockers including connexin mimetic peptides. These data allowed us to ascertain the functional and spatial organization of connexins at the MEJ.
Materials and Methods
An expanded Materials and Methods section is available in the online supplement at http://circres.ahajournals.org.
All procedures and protocols in this study were approved by the University of Virginia Animal Care and Use Committee. All mice (including Cx37−/− and Cx40−/−; kind gifts of D.L. Paul9,10) were on a C57Bl/6 background.
Isolation of mouse EC was based on methods described by Srinivasan et al 11 and Shi et al,12 with modifications. EC were used between passages 2 and 4. Isolation of VSMC was based on the method of Owens et al,13 with modifications. VSMC were used between passages 2 and 3.
A vascular cell coculture was created by inverting a FBS precoated polyester Transwell insert (Corning) and plating VSMC at 1×106 cells/insert. After 4 days, the insert was turned over and the upper membrane was plated with EC at 1×107 cells/insert. The coculture was allowed to stabilize for at least 72 hours.
Cell lysates were isolated with laemmli buffer, run on an SDS-PAGE gel, and transferred to polyvinylidene difluoride. Blots were developed using chemiluminescence (Pierce) and reprobed as needed (Chemicon).
Vascular cell cocultures were incubated for 30 minutes in supplemented PBS, primary antibodies (Abs) in supplemented PBS, washed with supplemented PBS, and incubated with corresponding secondary Abs in PBS. Samples were washed with PBS and imaged with a ×60 water immersion objective (0.90 NA) on an Olympus BX50WI confocal microscope under the control of Fluoview software.
To mark the cellular extensions into the Transwell, the membrane dyes FM 1-43FX and FM 4-64 (Molecular Probes) were applied to the VSMC or EC, respectively (30 μmol/L in Hanks’ balanced salt solution [HBSS[ with 25 mmol/L HEPES, 1% dimethyl sulfoxide).
EC were coloaded with biocytin (5 mg/mL; 357 Da, neutral; Molecular Probes) or Cy3 (1 mg/mL; 767 Da, Z=−1; Amersham Biosciences) and rhodamine-dextran (3 mg/mL; 4000 Da; Molecular Probes) or albumin coupled with fluorescein isothiocyanate (FITC) (3 mg/mL; >69 000 Da; Sigma) using a pinocytotic uptake method (Molecular Probes). Strepavidin 488 (1:1 binding biocytin; Molecular Probes) was used to detect biocytin in the EC and VSMC. Using Fluoview software, pores were identified with a ×100 water immersion objective and a measurement rectangle (0.8 μm×11 μm) was placed over a randomly selected pore that covered the entire focal plane of the pore. A Z-stack consisting of 4 successive XY sections in 0.2 μm steps was performed and then a computed image rotated 90 degrees (online Figure I). Using the reconstructed image, and starting at the basal side of the EC (x=0), the mean pixel intensity from XZ sections was obtained at 1-μm intervals. The pixel intensity at each point along the pore length was normalized to the maximum pixel intensity in the control conditions.
Gap Junction Inhibitors
Connexin-mimetic peptides are synthetic peptides corresponding to sequences on the extracellular loops of connexin proteins (for review, see Evans and Boitano14). Gap2737,43 or gap2740 (ADI and Sigma Genosys)7 were used at a concentration of 190 μmol/L in HBSS with 25 mmol/L HEPES. 18α–Glycyrrhetinic acid (18α-GA) (50 μmol/L; Sigma) was dissolved in 0.5% dimethyl sulfoxide in HBSS.
Cx43 short interference RNA (siRNA) duplexes (Qiagen; target sequence 1, 5-TAG AAG ATT CAA AGA GCT TAA-3; target sequence 2, 5-TCC CGT GGA GGT GGT ACT CAA-3; 25 nmol/L), as well as negative control siRNA (target sequence, 5-AAT TCT CCG AAC GTG TCA CGT-3; 25 nmol/L; Qiagen) were mixed with siLentFect (Bio-Rad) and transfected into EC or VSMC on the Transwell. None of the Cx43 target sequences affected total Cx37 or Cx40 protein expression (online Figure II).
Significance was at P<0.05 and determined by one-way ANOVA (Bonferroni post hoc test); error bars are ±SE.
Vascular Cell Coculture Model
Immunoblots demonstrated selective expression of desmin and α-smooth muscle actin in VSMC but not in EC (Figure 1A, 1B [lanes 1 and 2], and 1D). EC were positive for platelet-endothelial cell adhesion molecule (PECAM)-1 (Figure 2E and 2F [lanes 3 and 4]), whereas the VSMC were not (Figure 1F, lanes 1 and 2). EC also expressed VE-cadherin (Figure 1G), but not α-smooth muscle actin (Figure 1H).
In transverse sections (Figure 2A), EC and VSMC extended processes into the Transwell pores to make cell–cell contact (red box, Figure 2A and 2B). For clarification of cellular membranes at the contact points, the cell membranes of EC and VSMC were labeled with different lipophillic dyes (EC, FM 4-64, red; VSMC, FM 1-43FX, green). Red membrane was detectable ≈2.5 μm into the pore, and green fluorescence associated with the VSMC penetrated ≈7.5 μm into the pore (Figure 2C). Overlap was detectable over a distance of ≈5 μm (yellow fluorescence).
Connexin 37 was prominent in both EC and VSMC monolayers (Figure 3A and 3D) but never observed in the pores of the Transwell (Figure 3C). Cx40 was detectable only in the EC (Figure 3E and 3F). Cx43 was found throughout both EC and VSMC (Figure 3I through 3L). Immunoblots of EC (Figure 3M, lane 1) or VSMC (Figure 3M, lane 2) from the vascular cell coculture probed for Cx37, Cx40, and Cx43 confirmed the immunocytochemistry.
EC and VSMC reverse cultured on Transwells showed identical distribution of the connexins (Figure 4A), thus excluding orientation as a cue for the distribution of the connexins (Figure 4A1 through 4A3). When EC were grown on both sides of the Transwell insert (Figure 4B), Cx43, Cx40, and Cx37 were all detected in the cellular extensions (Figure 4B1 through 4B3). When VSMC were grown on both sides of the Transwell insert (Figure 4B), Cx43 and Cx37 were detected in the cellular extensions (Figure 4C1 through 4C2), and Cx40 was absent (Figure 4C3). Only in the case of EC–VSM contact was Cx37 absent from points of heterocellular contact.
Biocytin Dye Transfer
We assessed the functionality of the connexins by measuring diffusion of biocytin from EC into VSMC. En face views of the EC or VSMC on the vascular cell coculture demonstrated that biocytin could not diffuse to VSMC when 18α-GA was present but could in control conditions (Figure 5A). In an attempt to quantify cell–cell coupling at the MEJ, we examined biocytin distribution along the Transwell pore length. Biocytin was uniformly distributed along the length of the pore from EC to VSM (Figure 5B). Biocytin movement was inhibited by 18α-GA (Figure 5B). Connexin-mimetic peptides (Figure 5B; gap2740 and gap2737,43) each produced partial blockades, and their combination was additive (Figure 5B, combo), indicating that there was participation of more than one connexin isoform in the dye transfer.
To confirm our observations that Cx37 was present in EC, but not at the MEJ, Cx37−/− EC cultured from a Cx37−/− mouse were used. There was no effect on biocytin dye transfer to wild-type VSMC (Figure 5C), and 18α-GA was effective, as were gap2737,43 and gap2740 (Figure 5C). Similar dye transfer results were obtained from wild-type EC cocultured with Cx37−/− VSMC (online Figure IIIA). In addition, Cx37 did not redistribute to the MEJ when Cx37−/− EC or VSMC were used (online Figure IV).
When the vascular cell coculture was prepared with Cx40−/− EC and wild-type VSMC, dye transfer persisted, and the effect of 18α-GA was unaltered (Figure 5D), but gap2740 was ineffective (Figure 5D). Gap2737,43 and the combination of both peptides had identical effects on biocytin transfer. Biocytin movement from wild-type EC to Cx40−/− VSMC was similar to Figure 5A (online Figure IIIB), demonstrating no role for Cx40 in VSMC. Immunocytochemistry indicated that Cx37 did not redistribute to the MEJ when Cx40−/− EC were used (online Figure IV).
Effect of Cx43 siRNA on Biocytin Dye Transfer
Cx43 siRNA oligonucleotides transfected into EC or VSMC were able to significantly reduce Cx43 protein expression in both cell types. Immunoblots of EC and EC transfected with nonsilencing siRNA (Figure 6A, lanes 1 and 2) demonstrated Cx43 staining. When either Cx43 siRNA target sequence 1 (Figure 6A, lane 3) or Cx43 siRNA target sequence 2 (Figure 6A, lane 4) were transfected into EC, there was significant reduction in Cx43 protein. Untreated VSMC or those transfected with nonsilencing siRNA showed Cx43 staining (Figure 6A, lanes 5 and 6). Both of the Cx43 siRNA oligonucleotides reduced Cx43 protein in VSMC (Figure 6A, lanes 7 and 8).
Transfection of Cx43 siRNA into VSMC (Figure 6B) or EC (Figure 6C) resulted in significant Cx43 protein reduction (Figure 6B and 6C) in comparison with control sections (Figure 3J and 3K) and compared well with sections in which the Abs were preabsorbed with Cx43-GST peptide (Figure 6D15). Cx37 did not redistribute to the MEJ when Cx43 siRNA was present in EC or VSMC (online Figure IV).
We loaded biocytin into wild-type EC and examined transfer to VSMC transfected with Cx43 siRNA (Figure 7A). In control conditions, the diffusion of biocytin was halted 4 to 5 μm into the Transwell. Cocultures of wild-type EC and HeLa cells demonstrated biocytin dye transfer mimicked that seen in Figure 7A (online Figure IIIC).
Biocytin transfer persisted between EC with depleted Cx43 and wild-type VSMC (Figure 7B). Cx40−/− EC transfected with Cx43 siRNA showed reduced biocytin transfer to wild-type VSMC, similar to that seen in Figure 7A and online Figure IIIC. Taken together, our data indicate that EC–VSMC coupling is mediated by a combination of Cx40 and Cx43 in the EC and Cx43 in the VSMC. Dye transfer results with nonsilencing siRNA had no effect in either case (data not shown).
We also tested the intercellular diffusion of Cy3. Cy3 was impermeable between wild-type EC and VSMC (Figure 8A), but it diffused readily from Cx40−/− EC to VSMC (Figure 8B), and this movement was blocked by treatment with Cx43 siRNA (Figure 8C). Because of the likely connexin selectivity of the dye tracer,16,17 we believe this indicates that heterotypic gap junction were formed at the MEJ in wild-type conditions.
Myoendothelial junctions are thought to tightly regulate EC and VSMC communication; dye transfer,18 bidirectional signaling,6,7 and electrical coupling between the two cell types are well established.19,20 However, connexins involved and the signaling pathways activated are poorly understood,2 and, at this time, there is no information as to the connexins that establish this vital communication link. To address these issues, we have developed a vascular cell coculture system that is anatomically similar to that seen in vivo that produces cellular extensions that form functional MEJs.
In this model, Cx40 was expressed at points of EC–EC contact and at the MEJ, a distribution consistent with the prevalence of Cx40 in EC.21–23 Cx43 is also present in the EC, both at the perinuclear region and in the MEJ. We believe that Cx43 assembly in the ER/Golgi and expression in the EC extensions may be the reason the staining pattern in wild-type EC en face manifests a perinuclear stain. This distribution is different from what has been reported for the aorta in vivo, where Cx43 is present in EC on only the leading edge of arteriosclerotic plaques.24,25 The disparate observations may reflect that Cx43 in in vivo mouse EC is confined to a histologically inaccessible regions, such as the MEJ.2
We found Cx37 expressed at the pericellular boarders of the EC and diffusely expressed in VSMC (similar to in situ24,25), although, surprisingly, it was excluded from the point of EC and VSMC contact. In addition, Cx37 was found throughout the Transwell pores when both sides were seeded with either EC or VSMC (Figure 4), supporting the idea that heterocellular contact played a key role in guiding the organization of connexins, not just on connexin expression. Moreover, dye coupling measurements, and the use of selective blockers, combined with the alteration of protein expression, all confirmed the lack of Cx37 at the MEJ. In a similar example of Cx37 in heterocellular contact, Veitch et al have demonstrated that Cx37 protein was not present in granulosa cells until oocytes were present in the culture, implying oocytes were able to induce Cx37 upregulation in another cell type.26 In our system, Cx37 was always present in both cell types, either apart or in heterocellular contact, implying a possible role for heterocellular contact in control of connexin protein trafficking. It is clear that more work on Cx37 protein trafficking, as demonstrated in either system, is required.
The literature provides a variety of mechanisms for selective organization of connexin hexamers. Oligomerization of connexins could hypothetically occur at different locations in the ER/Golgi secretory pathway.27 It is possible that when heteromeric hexamers (such as the Cx40/Cx43 hexamer described here) are formed, they alter the location of oligomerization in the secretory pathway, which may lead directly to their transport to different parts of the cell.27,28 Thus, connexin hexamers could use different transport pathways for targeting to different regions of the plasma membrane, such as the MEJ.
We do not believe that connexins other than Cx40 and Cx43 are involved in EC and VSMC communication in this model. However, the mRNA for Cx45 has been detected in mouse EC,29 and, on Cx45 gene ablation from mouse, vessel formation is inhibited.30 In the vascular cell coculture model, Cx45 protein was not detected in the EC or the VSMC (data not shown). However, low-affinity Abs for this connexin have been reported.31 Based on the data from the knockout animals and the connexin-mimetic peptides (see below), we believe that if there are other connexins present in EC of VSMC, they behave similarly to Cx37.
Biocytin is a fixable molecule without net charge that diffuses through gap junctions with little or no selectivity.17 Using streptavidin coupled with Alexa 488 for detection ensured that fluorescence was not quenched during the fixation process. Transverse, Z-stack reconstruction of confocal images, rather than Z-sections from endothelium down to smooth muscle, enabled us to obtain unbleached pixel intensities through the entire length of the pore. In addition, this method of pore reconstruction reduces the impact of the axial-point spread function inherit in 3D reconstructions when compared with Z-stack reconstructions of pores from the en face view.
All gap junction inhibitors caused a decrease in fluorescence at the approximate point of MEJ formation within the pores at 4 to 5 μm (Figures 2 and 5⇑A). However, a key observation was that neither connexin-mimetic peptides alone produced a full blockade, although the combination led to a block that approximated that seen with 18α-GA. We interpreted these data as showing that multiple connexins existed at the MEJ. Gap2740 appeared to be specific for Cx40 as the peptide lost its effect when the connexin was absent from the preparation. Interestingly, gap2737,43 appeared not to be as effective an inhibitor as gap2740. The use of gap2737,43 should have produced a complete inhibition of dye transfer, assuming Cx43 was in both EC and VSMC. However, gap junction inhibitor experiments in other cell types also demonstrates that gap2737,43 may not be as potent as other connexin-mimetic peptides used to block Cx43.32 More research on the effect and mechanism of action of these peptides is required.
Immunocytochemistry initially suggested the presence of multiple connexins at the MEJ, including Cx40 and Cx43 in the EC and Cx43 in the VSMC. To confirm this, we examined biocytin transfer from Cx40−/− EC to wild-type VSMC. Deletion of Cx40 did not result in any changes in dye transfer to VSMC. However, when gap2740 was added, there was no inhibitory effect of the peptide. This suggested that Cx40 in EC may be a normal constituent of the gap junctions at the MEJ but that its function can be subsumed by Cx43 in its absence. The deletion of Cx43 from EC resulted in a similar pattern of dye transfer to that seen when wild-type EC and VSMC were used, indicating again that at least one connexin (likely Cx40) in the EC was still able to form gap junctions at the MEJ. In an attempt to determine exactly which connexins were present on the EC side of the MEJ, we added Cx43 siRNA to Cx40−/− EC. In these circumstances, all biocytin transfer from EC to VSMC was blocked. The only other instance of a complete block observed was when Cx43 protein was ablated from the VSMC (Figure 7A). These data are consistent with the overall concept that Cx40 and Cx43 from wild-type EC form functional gap junctions solely with Cx43 from wild-type VSMC.
Our data indicate that combinations of connexins form gap junctions at the MEJ. The possible connexin organizations included the following: (1) homotypic Cx43 gap junctions, (2) Cx40 hexamers from EC and Cx43 hexamers from VSMC forming heterotypic gap junctions, or (3) heteromeric hexamers of Cx40 and Cx43 from EC and hexamers of Cx43 from VSMC forming heterotypic gap junctions. We used cells from knockout mice to explore these possibilities. Elimination of Cx40 from the EC leaving only Cx43 on the EC side of the MEJ should have forced the formation of homotypic gap junctions between the two cells. We also deleted Cx43 from the EC with Cx43 siRNA, leaving only Cx40 from the EC in the MEJ, thereby forcing the formation of a heterotypic gap junction with the Cx43 from the VSMC. In each of these manipulations, biocytin dye transfer to VSMC indicated that functional gap junctional coupling remained at the MEJ (Figures 5C and 7⇑B).
The biophysical properties of biocytin would not likely distinguish between gap junctions composed of different connexins,16,17,33 and we, therefore, selected Cy3, which is similar to Alexa 594 (820 Da, Z=−1) and has been demonstrated to diffuse through Cx43 homotypic gap junctions but not heterotypic Cx40–Cx43 gap junctions.16 We assessed Cy3 diffusion from EC that were wild-type, Cx40−/−, or transfected with Cx43 siRNA (Figure 8) to wild-type VSMC. Only in Cx40−/− EC was Cy3 dye transfer to wild-type VSMC possible, ie, only in a situation in which it appeared that homotypic Cx43 gap junctions would have been present at the MEJ. The failure of Cy3 to diffuse from wild-type EC or Cx43 siRNA EC to VSMC argues that the functional gap junctions were, in this case, heterotypic. In view of the participation of Cx43 in functional coupling with VSMC when Cx40 is ablated in EC, heteromeric hexamers of Cx40 and Cx43 in EC seem to be the most likely combination that is formed at the junction.
In summary, our data suggest that heterocellular communication between VSMC and EC in this model of the MEJ is based on the spatial organization of the connexin distribution within the cells. The result is the formation of heterotypic gap junctions with a resultant charge or size selectively conferred on the transfer of solute between the two cell types (EC and VSMC). These data provide a strong rationale for future experiments in intact vessels. The expression of connexins forming heterotypic gap junctions has been observed in vivo,34 and studies on transformed cell lines and reconstituted hexamers have demonstrated that heterotypic gap junctions manifest differences in second messenger and dye transfer permeabilities.16,35,36 Therefore, we propose that spatial organization of connexins, and thus selective gap junction formation, may be used by EC and VSMC to tightly coordinate solute transfer between vascular cells and thus coordinate vascular function.
Supported by National Institutes of Health grants HL53318 (to B.R.D.), HL72864 (to B.R.D.), and HL79772 (to B.E.I.). We are grateful to K.H. Day, S.I. Ramos, X.F. Figueroa, and I. Rubio-Gayosso for expert comments. Transwell sections were from S. Vanhoose and J. Nash at the University of Virginia Research Histology Core. GST peptides were made with Dr R.K. Nakamoto and X. Liu. We thank S. Boitano for comments on the manuscript and G.J. Seedorf for culture ideas.
Original received March 29, 2005; revision received May 11, 2005; accepted June 3, 2005.
Sleek GE, Duling BR. Coordination of mural elements and myofilaments during arteriolar constriction. Circ Res. 1986; 59: 620–627.
Sandow SL, Hill CE. Incidence of myoendothelial gap junctions in the proximal and distal mesenteric arteries of the rat is suggestive of a role in endothelium-derived hyperpolarizing factor-mediated responses. Circ Res. 2000; 86: 341–346.
Yashiro Y, Duling BR. Participation of intracellular Ca2+ stores in arteriolar conducted responses. Am J Physiol Heart Circ Physiol. 2003; 285: H65–H73.
Chaytor AT, Martin PE, Edwards DH, Griffith TM. Gap junctional communication underpins EDHF-type relaxations evoked by ACh in the rat hepatic artery. Am J Physiol Heart Circ Physiol. 2001; 280: H2441–H2450.
Dora KA, Doyle MP, Duling BR. Elevation of intracellular calcium in smooth muscle causes endothelial cell generation of NO in arterioles. Proc Natl Acad Sci U S A. 1997; 94: 6529–6534.
Srinivasan S, Yeh M, Danziger EC, Hatley ME, Riggan AE, Leitinger N, Berliner JA, Hedrick CC. Glucose regulates monocyte adhesion through endothelial production of interleukin-8. Circ Res. 2003; 92: 371–377.
Shi W, Haberland ME, Jien ML, Shih DM, Lusis AJ. Endothelial responses to oxidized lipoproteins determine genetic susceptibility to atherosclerosis in mice. Circulation. 2000; 102: 75–81.
Owens GK, Loeb A, Gordon D, Thompson MM. Expression of smooth muscle-specific alpha-isoactin in cultured vascular smooth muscle cells: relationship between growth and cytodifferentiation. J Cell Biol. 1986; 102: 343–352.
Loo LW, Berestecky JM, Kanemitsu MY, Lau AF. pp60src-mediated phosphorylation of connexin 43, a gap junction protein. J Biol Chem. 1995; 270: 12751–12761.
Cottrell GT, Wu Y, Burt JM. Cx40 and Cx43 expression ratio influences heteromeric/ heterotypic gap junction channel properties. Am J Physiol Cell Physiol. 2002; 282: C1469–C1482.
Little TL, Xia J, Duling BR. Dye tracers define differential endothelial and smooth muscle coupling patterns within the arteriolar wall. Circ Res. 1995; 76: 498–504.
Figueroa XF, Paul DL, Simon AM, Goodenough DA, Day KH, Damon DN, Duling BR. Central role of connexin40 in the propagation of electrically activated vasodilation in mouse cremasteric arterioles in vivo. Circ Res. 2003; 92: 793–800.
Simon AM, McWhorter AR. Decreased intercellular dye-transfer and downregulation of non-ablated connexins in aortic endothelium deficient in connexin37 or connexin40. J Cell Sci. 2003; 116: 2223–2236.
Kwak BR, Veillard N, Pelli G, Mulhaupt F, James RW, Chanson M, Mach F. Reduced connexin43 expression inhibits atherosclerotic lesion formation in low-density lipoprotein receptor-deficient mice. Circulation. 2003; 107: 1033–1039.
Veitch GI, Gittens JE, Shao Q, Laird DW, Kidder GM. Selective assembly of connexin37 into heterocellular gap junctions at the oocyte/granulosa cell interface. J Cell Sci. 2004; 117: 2699–2707.
Martin PE, Blundell G, Ahmad S, Errington RJ, Evans WH. Multiple pathways in the trafficking and assembly of connexin 26, 32 and 43 into gap junction intercellular communication channels. J Cell Sci. 2001; 114: 3845–3855.
Hirschi KK, Burt JM, Hirschi KD, Dai C. Gap junction communication mediates transforming growth factor-beta activation and endothelial-induced mural cell differentiation. Circ Res. 2003; 93: 429–437.
Kruger O, Plum A, Kim JS, Winterhager E, Maxeiner S, Hallas G, Kirchhoff S, Traub O, Lamers WH, Willecke K. Defective vascular development in connexin 45-deficient mice. Development. 2000; 127: 4179–4193.
Boitano S, Evans WH. Connexin mimetic peptides reversibly inhibit Ca(2+) signaling through gap junctions in airway cells. Am J Physiol Lung Cell Mol Physiol. 2000; 279: L623–L630.
Altevogt BM, Paul DL. Four classes of intercellular channels between glial cells in the CNS. J Neurosci. 2004; 24: 4313–4323.
Valiunas V, Beyer EC, Brink PR. Cardiac gap junction channels show quantitative differences in selectivity. Circ Res. 2002; 91: 104–111.