Increase in Cx45 Gap Junction Channels in Cerebral Smooth Muscle Cells from SHR
We recently reported the novel finding of expression and function of connexin45 (Cx45) in cerebrovascular smooth muscle cells. We examined the hypothesis that Cx45 is altered in hypertension. Immunoblots for Cx45 showed a significant increase in Cx45 in cerebral arteries from adult spontaneously hypertensive rats (SHR) compared with adult Wistar-Kyoto (WKY) rats, with no difference in aorta or femoral artery. Patch-clamp of cerebral smooth muscle cells pairs from SHR versus WKY showed a significantly steeper voltage dependence of deactivation and partial block of junctional currents by quinine and by a peptide that interferes with docking of Cx45, consistent with dominance of functional Cx45 channels in SHR. We examined potential roles of blood pressure versus angiotensin in elevated Cx45 in SHR by measuring Cx45 protein in 4 groups: (1) long-term administration in Wistar rats of the nitric oxide synthase inhibitor L-NAME; (2) long-term administration in SHR of the ACE inhibitor captopril; (3) long-term administration in Wistar rats of angiotensin; and (4) exposure of basilar artery segments in organ culture to angiotensin. Blood pressure was significantly elevated in groups 1 and 3 and was normal in group 2. In groups 1, 2, and 4, there was no significant change in Cx45 protein. In group 3, there was a modest but insignificant increase in Cx45 protein but no change in voltage dependence of deactivation of junctional currents. Overall, our data show increased Cx45 in SHR that is unlikely to be due to either elevated blood pressure or to angiotensin. Relative dominance of Cx45 over Cx43 in cerebral vessels may predispose SHR to ischemic stroke.
Gap junctions are formed from an assembly of transmembrane channels that are dodecamers of proteins of the connexin (Cx) family. Depending on which specific connexins are involved, these channels may allow direct exchange of ions or small signaling molecules between adjacent cells. For the endothelial and smooth muscle cells (SMC) of the vessel wall, gap junctions are formed from Cx37, Cx40, Cx43, and Cx45, each of which have electrical and conductive properties that are different from one another.1–3⇓⇓ When compared with Cx43, homotypic Cx45 channels exhibit a small conductance, low permeability to Ca2+, anions, and Lucifer yellow and are likely to contribute much less to metabolic coupling.4–6⇓⇓ Thus, the relative proportion of each connexin-forming gap junction is important for determining functional connectivity.
Surprisingly, little has been found of changes in connexins associated with hypertension. In two models of acquired hypertension, the DOCA-salt model and the 2-kidney, 1-clip Goldblatt model, Cx43 protein and mRNA are increased in aorta, whereas in hypertension induced by inhibition of nitric oxide synthase, Cx43 is decreased in aorta.7–10⇓⇓⇓ Findings in genetic hypertension are more limited. In mesenteric arteries from stroke-prone spontaneously hypertensive rats, the amount of Cx43 mRNA is similar to that in control Wistar-Kyoto (WKY) rats,11 whereas endothelium from tail arteries of spontaneously hypertensive rats (SHR) shows a significant decrease in Cx40, Cx43, and Cx37.12 Thus, to date, reports of alterations of connexins in hypertension have been limited, and specific findings have depended on the particular model under study.
All of the work on connexins in hypertension has focused on Cx43, Cx40, and Cx37, the three most widely recognized connexins in blood vessels. Recently, we3 and others2,13⇓ discovered that Cx45 is also a prominent member of the connexin family in blood vessels. In the current study, we sought to determine whether Cx45 might be altered in hypertension. We studied vessels from normotensive rats and from three models of hypertension, SHR, hypertension induced by long-term administration of NG-nitro-l-arginine methyl ester (L-NAME), and hypertension induced by long-term administration of angiotensin-II (Ang), ie, Ang hypertensive rat (AHR). We compared expression of Cx45 and Cx43 protein, and we measured electrophysiological properties in SMC pairs. Our data reveal that Cx45 expression was significantly increased in cerebral but not in noncerebral vessels from SHR and that this led to demonstrable alterations in properties of electrical coupling between SMC. Our data also indicate that acquired hypertension, induced by either L-NAME or Ang, did not lead to similar changes, suggesting that neither elevated blood pressure nor Ang can be directly implicated in the increase in Cx45 in SHR.
All studies were conducted with female rats. SHR and their genetic controls, WKY rats, were obtained from Charles River Laboratory. For L-NAME–induced hypertension and for AHR, Wistar rats (250 g) were implanted subcutaneously with an ALZET osmotic pump (Durect Corporation) filled either with L-NAME (Sigma), delivered at 200 mg/kg per day for 4 weeks,14 or with Ang (Sigma), delivered at 265 μg/kg per day for 4 weeks.15 Anesthesia for survival surgery was obtained by intraperitoneal administration of ketamine (60 mg/kg) plus zylazine (8 mg/kg). Blood pressure was monitored by means of tail-cuff plethysmography.
Smooth Muscle Cells
Basilar and posterior cerebral arteries were dissected from WKY rats (16 to 20 weeks), SHR (16 to 20 weeks), Wistar rats (250 to 300 g), and AHR (12 to 14 weeks). SMC pairs were isolated by enzymatic digestion16 and used for electrophysiological experiments within 24 hours. In some preparations of SHR SMC, cerebral arteries were isolated in the presence of anti-Cx45 peptide.3
Cerebral arteries were pooled from 3 to 6 age-matched rats in each of the following 9 groups: young WKY rats (5 weeks); adult WKY rats (18 weeks); young SHR (5 weeks); adult SHR (18 weeks); adult SHR treated with captopril (10 mg/kg per day for 4 weeks, through osmotic pump), which were normotensive; adult Wistar rats (12 weeks); adult AHR; adult Wistar rats with L-NAME–induced hypertension; adult Wistar rats treated with captopril (10 mg/kg per day for 4 weeks, through osmotic pump). For some experiments, cerebral vessels were harvested from adult Wistar rats and segments were incubated with Ang (0, 30 nmol/L, 300 nmol/L) at 37°C for 6 hours before further processing. Aortas and femoral arteries were pooled from 4 adult WKY rats and SHR.
Tissues were homogenized in lysate buffer containing 10 mmol/L Tris-HCl, 1 mmol/L EDTA, 1% Nonidet-P40, 250 mmol/L sucrose, 1% protease inhibitor cocktail (Sigma), 1% phosphatase inhibitor cocktail I and II (Sigma), and 1% 2-mercaptoethanol, pH 6.8. Subconfluent HeLa and F9 cells were lysed in RIPA buffer. Aliquots (40 μg total protein) were separated on SDS-PAGE and transferred to PVDF membranes. Cx43 was probed with mouse anti-Cx43 monoclonal antibody (1:1000, Chemicon); Cx45 was probed with rabbit anti-Cx45 polyclonal antibody (1:5000, generously provided by Dr T.H. Steinberg, Washington University, St Louis); β-actin was probed with mouse anti–β-actin monoclonal antibody (1:5000, Sigma). Probes were labeled with horseradish peroxidase–labeled species-appropriate IgG (1:1000 or 1:5000, Amersham) and were detected by chemiluminescence (ECL, Amersham). Autoradiographs were scanned and quantified by densitometry (Scion Image).
Immunofluorescence Labeling and Electrophysiology
Immunolabeling of vessel segments and isolated cells as well as dual perforated patch-clamp experiments were carried out as previously described.3
Data were fit to Boltzmann3 or gaussian functions by using Origin 7.0 (Microcal Software, Inc). Data are given as mean±SEM except where otherwise indicated. Statistical significance was assessed by using ANOVA, t test, or χ2 analysis.
Spontaneously Hypertensive Rats
Initial experiments were aimed at determining expression of Cx45 protein in blood vessels from adult WKY and SHR. We studied cerebral (basilar and posterior cerebral) arteries and compared these to similarly sized femoral arteries as well as aorta, with HeLa and F9 cells serving as negative and positive controls,3 respectively. Cx45 was visibly more prominent in cerebral vessels from SHR compared with WKY (Figure 1A). After quantification by densitometry and normalization to levels of β-actin, Cx45 protein was found to be significantly elevated (by ANOVA, P<0.05) in cerebral vessels from SHR compared with WKY rats (Figure 1B). By comparison, levels of Cx45 were similar in aorta and femoral artery of SHR compared with WKY (Figures 1A and 1B).
Immunofluorescence study of cerebral vessels indicated abundant labeling for Cx45 in basilar and posterior cerebral arteries from SHR (Figure 2A). Cx45 labeling was confined to smooth muscle layers, with no labeling of endothelial or adventitial layers. By comparison, Cx43 was present in medial layers of SHR but was less prominent than Cx45, and Cx43 was apparent in both endothelial and adventitial layers (Figure 2B). Nuclear labeling with DAPI was used to confirm the identity of cell layers (Figure 2C).
Double labeling of freshly dissociated pairs of SMC from SHR revealed that Cx45 and Cx43 could be coexpressed in the same cells, although in different patterns. Cx45 labeling was arranged linearly along SMC membranes and was often particularly prominent at cell-to-cell interfaces (Figure 2D). Cx43 tended to be more punctate in appearance and was more diffusely distributed (Figure 2E). DAPI labeling of nuclei confirmed the presence of two cells (Figure 2F).
In additional separate experiments, we confirmed increased expression of Cx45 protein in cerebral vessels from SHR compared with WKY (Figures 3A and 3B). Cx43 was also present in tissue from cerebral vessels (Figure 3A), but no significant difference in level of expression was found in WKY versus SHR (Figure 3C).
We also examined cerebral vessels from 5-week-old WKY versus SHR, with the latter not yet hypertensive. In both groups, levels of Cx45 were comparable to adult SHR (Figure 3B). Similarly, in both groups, levels of Cx43 were similar to those in adults (Figure 3C).
We next determined whether the increase in Cx45 protein found in cerebral vessels from adult SHR would be reflected in altered electrophysiological properties of gap junctions in isolated SMC pairs. In Wistar rats, we previously found that one third (10 of 26) of cell pairs exhibited junctional conductances with rapid deactivation and steep voltage-dependence consistent with dominance of Cx45 channels, with the remaining pairs showing slower deactivation and weaker voltage dependence consistent with dominance of Cx43 or Cx40 channels.3 Here, we carried out a similar analysis of junctional currents in WKY versus SHR. Original records of junctional current obtained in a cell pair from SHR exhibiting rapid deactivation are shown in Figure 4A. At low transjunctional potentials, little deactivation was apparent, but as transjunctional potential increased beyond 10 mV, strong, rapid deactivation was evident. A plot of the normalized steady-state (6-second) conductance, G′j-ss as a function of junctional potential, Vj, is shown for this cell pair (Figure 4B). Similar data obtained from 31 and 46 cell pairs from WKY rats and from SHR, respectively, are plotted in Figures 4C and 4D, respectively. Although both groups contained pairs with strong and weak voltage dependence of deactivation, pairs with strong voltage dependence of deactivation dominated the SHR group. Overall, a significantly different proportion (by χ2, P<0.01) of pairs, 13 of 31 versus 33 of 46 from WKY versus SHR rats, respectively, showed sufficiently strong voltage dependence that 55% or less of the steady-state current remained at +30 mV.
Based on our previous work,3 we considered that the strongly voltage-dependent deactivation evident in many cell pairs from SHR was likely to be due to Cx45. To evaluate this, we tested the effect of quinine, which has been shown to partially but selectively block junctional currents caused by Cx45.17 At concentrations of 0.3 mmol/L and 1 mmol/L, quinine reversibly blocked up to half of the junctional conductance in pairs from SHR exhibiting strong voltage-dependent deactivation (Figures 5A and 5B). We also tested the effect of an oligopeptide that selectively interferes with docking of Cx45 hemichannels.3 Treatment with anti-Cx45 peptide significantly decreased (by t test, P<0.05) the macroscopic junctional conductance (Figure 5C) and led to weaker voltage dependence of deactivation (Figure 5D). In SHR pairs treated with anti-Cx45 peptide, only 5 of 15 pairs showed sufficiently strong voltage dependence that 55% or less of the steady-state current remained at +30 mV, a proportion that was significantly different (by χ2, P<0.02) from 33 of 46 in untreated pairs. Together, these data showing strong voltage dependence of deactivation, combined with sensitivity to quinine and to anti-Cx45 peptide, provided evidence that the Cx45 protein identified in Western blots was forming functional gap junction channels in SMC pairs from SHR.
The significant augmentation of Cx45 protein and of functional channels in cerebral SMC pairs from SHR raised the possibility that elevated blood pressure could be responsible for increasing Cx45. To assess this, we measured Cx45 after long-term administration of the nitric oxide synthase inhibitor L-NAME in Wistar rats.14 These animals responded to long-term L-NAME administration by elevation of systolic blood pressure (BPsys) to values (mean±SD) of 237±6 mm Hg, compared with values of 141±12 mm Hg in untreated animals. However, levels of Cx45 protein, as well as those of Cx43 protein, were not appreciably changed in cerebral vessels (Figures 6A through 6C), suggesting that high blood pressure per se was not the cause of augmented Cx45 in SHR.
Hypertension in SHR may be treated successfully with the use of Ang ACE inhibitors such as captopril,18 suggesting involvement of Ang in the pathophysiology of hypertension in SHR. We thus sought to determine whether Ang could be implicated in augmented Cx45 in cerebral vessels of SHR. We examined Cx45 protein in two in vivo models relevant to Ang: (1) in SHR after long-term administration of captopril; and (2) in Wistar rats after long-term administration of subpressor Ang (AHR).
In SHR treated with captopril, values (mean±SD) of BPsys were reduced to 134±8 mm Hg, compared with values of 191±14 mm Hg in untreated SHR. However, levels of Cx45 protein, as well as those of Cx43 protein, were not significantly changed in cerebral vessels (Figures 3A through 3C). Captopril significantly reduced (by ANOVA, P=0.05) Cx43 in cerebral vessels from Wistar rats (Figures 6A and 6C), but this was not investigated further.
In AHR, values (mean±SD) of BPsys were elevated to 225±32 mm Hg, compared with values of 141±12 mm Hg in untreated Wistar rats. In these animals, a modest increase in both Cx45 and Cx43 was found in cerebral vessels, but these increases were not statistically significant (by ANOVA, P=0.09; Figures 6A through 6C).
To further examine the possible relation between Ang and Cx45, we measured the electrophysiological properties of gap junction channels in isolated SMC pairs. Data on the G′j-ss-Vj, relation obtained from 45 and 47 cell pairs from Wistar rats and from AHR, respectively, are plotted in Figures 6D and 6E, respectively. Both groups contained pairs with strong as well as weak voltage dependence of deactivation. Overall, 15 of 45 versus 18 of 47 pairs from Wistar rats versus AHR, respectively, showed sufficiently strong voltage dependence that 55% or less of the steady-state current remained at +30 mV. These proportions (15 of 45 and 18 of 47) were not significantly different (by χ2, P>0.05) and were similar to our previously published observations in Wistar rats (10 of 26).3 Thus, despite the tendency to exhibit more Cx45 as well as Cx43 protein in AHR, these changes were not reflected in any apparent change in electrical connectivity between isolated cell pairs.
Given the suggestion above of an elevation in Cx45 in AHR, we also measured Cx45 expression in isolated basilar arteries exposed to Ang for 6 hours in vitro. In these experiments, Ang did not appreciably alter Cx45 at concentrations up to 300 mmol/L (by ANOVA, P>0.05; not shown).
Strong Versus Weak Voltage Dependence of Deactivation
The electrophysiological data presented above were obtained from a total of 169 cell pairs. This large number of observations allowed us to assess the validity of a criterion we previously established for assessing gap junctional connectivity in SMC pairs. In our previous report,3 we cited the proportion of cell pairs “showing sufficiently strong voltage dependence that half or more of the current is deactivated at +30 mV.” This simple measure appeared to correlate well with junctions dominated by Cx45 versus those dominated by Cx43 and Cx40. Here, we compiled a frequency histogram of the normalized steady-state junctional conductance at +30 mV (G′j-30 mV) in the 169 pairs pooled from WKY, SHR, Wistar rats, and AHR (Figure 7A). The data were distributed into two groups, and the histogram was well fit with the sum of two gaussian functions, with values (mean±SD) of 0.29±0.13 and 0.96±0.05. Thus, a discriminating cutoff, the mean plus 2 SD, would be 0.86 for the group with weak voltage dependence and 0.55 for the group with strong voltage dependence, with the latter agreeing with the value of 0.5 that we previously used. With the use of these criteria to segregate observations into two groups, those with strong or weak voltage dependence of deactivation, SHR was found to have a significantly higher (by χ2, P<0.01) proportion of pairs exhibiting strong voltage dependence, compared with WKY, Wistar, and AHR (Figure 7B).
The principal finding of this study was that Cx45 was markedly increased in cerebral but not noncerebral vessels from SHR compared with WKY, resulting in a significant alteration in electrical connectivity between cerebral SMC. In cerebral vascular SMC from Wistar rats, we previously showed coexistence of multiple functional connexin channels, including homotypic Cx43, Cx45, and Cx40 channels, as well as nonhomotypic channels, which together lead to complex connectivity between cells.3,16⇓ Similarly, complex connectivity caused by expression of the same connexins may exist in vessels from SHR, but the relative dominance of Cx45 in SHR resulted in a different electrophysiological picture.
The signature finding of gap junctions dominated by Cx45 was strong voltage dependence of deactivation. We used the electrophysiological criterion of 55% or less of the steady-state current remaining at +30 mV to discriminate between junctions dominated by Cx45 and those dominated by Cx43 or Cx40. This approach was justified by our previous3 as well as present experiments using anti-Cx45 peptide and quinine17 and was further substantiated by our observation of two distinct populations of SMC pairs, based on an analysis of recordings from 169 pairs. This electrophysiological characteristic appears to be a robust correlate of Cx45 dominance.
No alteration of any specific connexin has ever been linked with hypertension. Expression of Cx43 may be either increased or decreased, depending on the specific model,7–10⇓⇓⇓ suggesting that blood pressure per se is not a direct regulatory factor for Cx43. Based on our findings in this study, the same appears to be true for Cx45. For Cx43, expression is linked to the phenotypic or proliferative status of SMC,19,20⇓ but whether this is true for Cx45 remains to be determined.
Our data indicating that Ang was not a regulator of Cx43 or Cx45 in SMC were surprising, given that Ang is known to upregulate Cx43 synthesis in rat ventricular myocytes.21,22⇓ Although mechanisms regulating Cx45 expression are not well understood, this connexin is known to be commonly downregulated after early development. In cardiac and motor neurons, maturation is associated with reduced expression of Cx45.23,24⇓ In the vascular system, Cx45 expression is critical during early development,13 but any decline in expression after development had not been examined. Our data indicated that Cx45 levels were elevated in cerebral vessels from both WKY rats and SHR at 5 weeks and that levels diminished with maturation in WKY rats but not SHR, suggesting that elevated levels of Cx45 in adult SHR may represent failure of normal developmental downregulation.
The relative dominance of Cx45 in SHR is expected to influence the nature of the signals transmitted between cells. Cx43 and Cx45 form junctional channels with different molecular permeabilities, and thus the consequence of a shift from Cx43 dominance to Cx45-dominance is likely to be significant. Homotypic Cx43 channels are highly permeable to cations and anions, including Ca2+ and the divalent anion Lucifer yellow, consistent with an ability to mediate intercellular exchange of anionic molecules such as cAMP, cGMP, and IP3.4,6⇓ Conversely, homotypic Cx45 channels exhibit a small conductance, low permeability to Ca2+, anions, and dye molecules, and thus probably contribute much less to metabolic coupling.4–6⇓⇓
Whether the difference that we observed in Cx45 in SHR versus WKY can account for any specific aspect of altered function in cerebral vessels remains to be determined, but a potentially relevant functional corollary may be the susceptibility of SHR to ischemic stroke.25,26⇓ Patency and opening of collateral vessels are crucial determinants of infarct size.27,28⇓ In SHR, the increase in susceptibility to infarction is attributed to a lack of collateral flow and narrower anastomoses.29,30⇓ Cx43 gap junctions, with their weak voltage dependence and nondiscriminating permeability, are critical for enhancing opening of collaterals, as shown by the fact that Cx43-null mice exhibit an exaggerated susceptibility to focal cerebral ischemia.31 Thus, Cx45 dominance, as in SHR, may result in the functional equivalence of a deficiency in Cx43, impairing the opening of collateral channels and thereby worsening cerebral ischemia.
In this study, we showed that cerebral vessels from SHR express greater levels of functional Cx45 compared with noncerebral vessels and compared with cerebral vessels from their genetic controls, WKY rats. Moreover, our data indicate that this is likely to be due to a genetic defect rather than to hypertension per se. This is the first demonstration of a difference in gap junction channel expression and function in cerebral vessels from hypertensive animals. Given the known differences in permeability between Cx45 and Cx43, we speculate that this finding may be an important correlate of the tendency of these animals to have greater damage from cerebral ischemia.
This work was supported by grants (to J.M.S.) from the National Heart, Lung, and Blood Institute (HL51932), the National Institute for Neurological Diseases and Stroke (NS39956), and a Bugher award from the American Heart Association.
- Received June 7, 2002.
- Revision received July 9, 2002.
- Accepted October 4, 2002.
Christ GJ, Spray DC, el Sabban M, Moore LK, Brink PR. Gap junctions in vascular tissues: evaluating the role of intercellular communication in the modulation of vasomotor tone. Circ Res. 1996; 79: 631–646.
Ko YS, Coppen SR, Dupont E, Rothery S, Severs NJ. Regional differentiation of desmin, connexin43, and connexin45 expression patterns in rat aortic smooth muscle. Arterioscler Thromb Vasc Biol. 2001; 21: 355–364.
Li X, Simard JM. Connexin45 gap junction channels in rat cerebral vascular smooth muscle cells. Am J Physiol Heart Circ Physiol. 2001; 281: H1890–H1898.
Veenstra RD, Wang HZ, Beyer EC, Brink PR. Selective dye and ionic permeability of gap junction channels formed by connexin45. Circ Res. 1994; 75: 483–490.
Jorgensen NR, Geist ST, Civitelli R, Steinberg TH. ATP- and gap junction-dependent intercellular calcium signaling in osteoblastic cells. J Cell Biol. 1997; 139: 497–506.
Watts SW, Webb RC. Vascular gap junctional communication is increased in mineralocorticoid-salt hypertension. Hypertension. 1996; 28: 888–893.
Haefliger JA, Castillo E, Waeber G, Bergonzelli GE, Aubert JF, Sutter E, Nicod P, Waeber B, Meda P. Hypertension increases connexin43 in a tissue-specific manner. Circulation. 1997; 95: 1007–1014.
Haefliger JA, Meda P, Formenton A, Wiesel P, Zanchi A, Brunner HR, Nicod P, Hayoz D. Aortic connexin43 is decreased during hypertension induced by inhibition of nitric oxide synthase. Arterioscler Thromb Vasc Biol. 1999; 19: 1615–1622.
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.
Ribeiro MO, Antunes E, de Nucci G, Lovisolo SM, Zatz R. Chronic inhibition of nitric oxide synthesis: a new model of arterial hypertension. Hypertension. 1992; 20: 298–303.
Simon G, Abraham G, Cserep G. Pressor and subpressor angiotensin II administration. Two experimental models of hypertension. Am J Hypertens. 1995; 8: 645–650.
Li X, Simard JM. Multiple connexins form gap junction channels in rat basilar artery smooth muscle cells. Circ Res. 1999; 84: 1277–1284.
Srinivas M, Hopperstad MG, Spray DC. Quinine blocks specific gap junction channel subtypes. Proc Natl Acad Sci U S A. 2001; 98: 10942–10947.
Ito K, Koike H, Miyamoto M, Ozaki H, Kishimoto T, Urakawa N. Long-term effects of captopril on cellular sodium content and mechanical properties of aortic smooth muscle from spontaneously hypertensive rats. J Pharmacol Exp Ther. 1981; 219: 520–525.
Polontchouk L, Ebelt B, Jackels M, Dhein S. Chronic effects of endothelin 1 and angiotensin II on gap junctions and intercellular communication in cardiac cells. FASEB J. 2002; 16: 87–89.
Chang Q, Gonzalez M, Pinter MJ, Balice-Gordon RJ. Gap junctional coupling and patterns of connexin expression among neonatal rat lumbar spinal motor neurons. J Neurosci. 1999; 19: 10813–10828.
Alcolea S, Theveniau-Ruissy M, Jarry-Guichard T, Marics I, Tzouanacou E, Chauvin JP, Briand JP, Moorman AF, Lamers WH, Gros DB. Downregulation of connexin 45 gene products during mouse heart development. Circ Res. 1999; 84: 1365–1379.
Coyle P, Heistad DD. Blood flow through cerebral collateral vessels one month after middle cerebral artery occlusion. Stroke. 1987; 18: 407–411.
Coyle P. Spatial relations of dorsal anastomoses and lesion border after middle cerebral artery occlusion. Stroke. 1987; 18: 1133–1140.