Potential conflict of interest: Nothing to report.
This study was supported by Action Medical Research (SP-4219 to J.G.an C.W.), the Wellcome Trust (WTN09594 to J.G.), BHF Grant (NH/10/3/28574 to JG), the Biomedical Research Center at Imperial College Healthcare National Health Service Trust, and by the Swiss National Science Foundation (3100AO-105916 to S.R.).
Intrahepatic cholestasis of pregnancy may be complicated by fetal arrhythmia, fetal hypoxia, preterm labor, and, in severe cases, intrauterine death. The precise etiology of fetal death is not known. However, taurocholate has been demonstrated to cause arrhythmia and abnormal calcium dynamics in cardiomyocytes. To identify the underlying reason for increased susceptibility of fetal cardiomyocytes to arrhythmia, we studied myofibroblasts (MFBs), which appear during structural remodeling of the adult diseased heart. In vitro, they depolarize rat cardiomyocytes via heterocellular gap junctional coupling. Recently, it has been hypothesized that ventricular MFBs might appear in the developing human heart, triggered by physiological fetal hypoxia. However, their presence in the fetal heart (FH) and their proarrhythmogenic effects have not been systematically characterized. Immunohistochemistry demonstrated that ventricular MFBs transiently appear in the human FH during gestation. We established two in vitro models of the maternal heart (MH) and FH, both exposed to increasing doses of taurocholate. The MH model consisted of confluent strands of rat cardiomyocytes, whereas for the FH model, we added cardiac MFBs on top of cardiomyocytes. Taurocholate in the FH model, but not in the MH model, slowed conduction velocity from 19 to 9 cm/s, induced early after depolarizations, and resulted in sustained re-entrant arrhythmias. These arrhythmic events were prevented by ursodeoxycholic acid, which hyperpolarized MFB membrane potential by modulating potassium conductance. Conclusion: These results illustrate that the appearance of MFBs in the FH may contribute to arrhythmias. The above-described mechanism represents a new therapeutic approach for cardiac arrhythmias at the level of MFB. (HEPATOLOGY 2011;)
Intrahepatic cholestasis of pregnancy (ICP) is a common disorder presenting in the third trimester of pregnancy.1 Fetal complications include fetal anoxia, meconium-stained liquor, and preterm labor, all occurring more commonly in pregnancies with higher maternal serum bile acids.2 ICP is also associated with fetal arrhythmia (FA)3, 4 and intrauterine death.1, 5 Fetal death in ICP affects 2%-4% of ICP pregnancies.6 Interestingly, fetal atrial flutter and supraventricular tachycardia occur in ICP during the second and the third trimester.3, 4 These FAs may be related to alterations within the cardiac conduction system when seen in pregnancies complicated by maternal disease.6 Increases in plasma levels of bile acids, in particular taurocholic acid (TC), which has been demonstrated to cross the placental barrier, are probably produced by pathology.1 We have described the mechanism of TC-induced FA, which was associated with a desynchronization of calcium-ion (Ca2+) dynamics in cultured neonatal rat cardiomyocytes (CMs), treated with TC that acts as a partial agonist at the muscarinic M2 receptors.7 Although fetal complications have been reported in ICP, there are no reports of maternal arrhythmia in affected women. It is still unknown why TC-induced arrhythmia occurs only in the fetal heart (FH) and not in the maternal heart (MH).
The human heart develops in a partially hypoxic environment, which is necessary for the normal development of the vasculature and other systems.8 Moreover, hypoxia is thought to contribute to the in vitro transdifferentiation of human fetal cardiac fibroblasts into myofibroblasts (MFBs).9 Interstitial MFBs also appear in ischemic, fibrotic, and infarcted adult hearts.10, 11 They are thought to contribute to arrhythmogenesis by excess deposition of extracellular matrix, leading to disruption of the orderly three-dimensional (3D) network of electrically coupled CMs. Besides their participation in the structural remodeling of diseased hearts, we have recently shown, in vitro, that MFBs are capable of establishing arrhythmogenic heterocellular gap junctional coupling with CMs, which results in the modulation of CM electrophysiology.12–14 Therefore, the transient appearance of MFBs in FH might provide a substrate for arrhythmia.
We show here that MFBs transiently appear in human fetal ventricular tissue during the second and the third trimester of gestation, when cholestasis-related sudden fetal death is most common. We, therefore, explored whether TC would induce arrhythmia by investigating cardiac impulse propagation characteristics using a previously described model,13 consisting of a heterocellular preparation of a confluent monolayer of CMs coated with a layer of MFBs. For simplicity, this was named the “FH model.” Apart from the observation that TC induces arrhythmias under these conditions, we found that ursodeoxycholic acid (UDCA), a commonly used drug to treat ICP, protects against the arrhythmogenic effects of TC by directly affecting the membrane potential (Vm) of MFBs. These findings suggest that MFBs might be involved in triggering fetal ICP-induced arrhythmia, and that these cells possibly represent a new specific target for antiarrhythmic therapy.
An expanded Methods section is available in the Supporting Information.
Patterned Growth Cell Cultures of Neonatal Rat Ventricular Cardiomyocytes.
Isolation procedures from 1-day-old Wistar rats were in accord with the guidelines of the UK Home Office Animal (Scientific Procedures) Act of 1986.
For details, see the Supporting Information.
Human fetal ventricular tissue slices obtained from the Fetal Tissue Bank (Hammersmith, Hospital, London), the postnatal human sample (1 day old) and the adult human infarct (4 months old) from the Department of Histopathology, Royal Brompton Hospital (London, UK), were stained for alpha smooth muscle actin (α-SMA; Thermo Scientific, Dako, Fremont, CA), vimentin (Invitrogen, Carlsbad, CA), and discoidin domain receptor 2 (DDR2; R&D Systems, Inc., Minneapolis, MN). The conventional horseradish-peroxidase technique was used for immunohistochemical investigations.
Optical and Electrical Measurement of Action Potential Characteristics.
Microscopic measurements of impulse propagation characteristics of the conducted action potentials (APs) were performed using a custom-made optical recording system.15 Macroscopic measurements of electrical activity were performed using a custom-built macroscope, in conjunction with a fast CMOS camera (MiCAM ULTIMA; SciMedia USA Ltd., Costa Mesa, CA).14Vm values in the monolayer of (1) CMs, (2) CM-MFBs, and (3) pure MFBs were measured at 36°C using sharp electrode impalements in selected regions of the cell, controlled by a scanning ion conductance microscope (SICM).16, 17 Bile acids and glibenclamide were from Sigma-Aldrich (St. Louis, MO).
Data obtained with the macroscopic imaging system were analyzed using proprietary software (ULTIMA; SciMedia USA). Microscopic impulse propagation data were analyzed using custom-made software after low-pass filtering of the signals (f0 = 0.5 kHz). Optically measured action potential amplitudes (APAs) were set to 100% (%APA), and maximal upstroke velocities were calculated accordingly (%APA/ms). With an assumed APA of 100 mV, %APA/ms corresponds to V/s. Transmembrane potentials were averaged over 1 minute. Based on previous experimental protocols,13 only impalements exhibiting stable potentials for 60-70 seconds were analyzed (Clampfit 10.0; Molecular Devices,Inc., Sunnyvale, CA).
Data differences were considered significant at P < 0.05 (Student's t-test, two-tailed, heteroscedastic).
Transient Appearance of MFBs in Human FH Tissue During Development.
We analyzed human ventricular samples derived from fetuses at 9-26 weeks of gestation to characterize the presence of MFBs in FH during development. Slices were stained for α-SMA (Fig. 1), vimentin, and discoidin-domain 2 receptor (Supporting Fig. 1A,B). Figure 1A shows modest detectable α-SMA expression in the first trimester of gestation. However, during the second and the third trimester, α-SMA-positive areas became more defined, showing broader distribution, resulting in a close contact between MFBs and resident CMs. As expected, no MFBs could be detected in postnatal samples, apart from α-SMA positivity within the blood vessels. Quantifying the density of MFBs at different stages of gestation, Fig. 1B illustrates that the maximal density of MFBs rises to 17.1% at 15 weeks and, with advancing gestation, continues ranging between 17.1% and 13.2%. As a positive control, we analyzed a section of (1) 6-week-old human healing granulation tissue from a ventricular infarcted area (Fig. 1C) and (2) chronically infarcted area induced in the rat heart (Supporting Fig. 1). Both infarct regions are characterized by robust α-SMA-positive staining, suggesting the presence of numerous MFBs.
Effects of TC on Impulse Conduction in the MH and FH Models.
Next, we studied the effects of TC using in vitro models of the MH and FH. The healthy adult heart is normally devoid of MFBs. To model this situation in vitro, we constructed homocellular strands of CMs,14 mostly devoid of MFBs (Fig. 2; left). For FHs, we added a layer of cardiac MFBs on top of the confluent CM strands that formed a uniform dense coat after 2 days in culture (Fig. 2; right). ICP was simulated by exposing MHs and FHs to TC at concentrations typical of severe ICP (0.5 mmol/L, 15 minutes acute, 12-16 hours chronic; see Supporting Information for case report).7 Conduction velocity (θ) and maximal upstroke velocities (dV/dtmax) were high in the MH model under control conditions (37.1 cm/s; 82.1 %APA/ms) and in the presence of 0.5 mmol/L of TC (32.1 cm/s; 81.8 %APA/ms). In contrast, the FH model θ (19.8 cm/s) and dV/dtmax were low (43.9 %APA/ms), as previously shown in this heterocellular preparation.13 Acute administration of TC to the FH model markedly decreased θ from 19.8 to 9.2 cm/s (Fig. 2; lower row, right column) and reduced dV/dtmax from 43.9 to 28 %APA/ms. Figure 2B shows a summary of the conduction velocity in the MH and FH models during both acute (10-20 minutes) and chronic (12-16 hours) administration of TC. In the maternal model consisting predominantly of CMs, neither acute nor chronic exposure to TC had any effects on θ (Fig. 2B; left). By contrast, acute administration of TC to the FH model reduced θ dose dependently by ∼40% and, with chronic administration, by ∼52% (Fig. 2B; right).
In addition, we tested the hypothesis that elevated TC might affect automaticity in the FH model (consisting of a 1-cm-diameter, disc-shaped isotropic monolayer of CMs coated with MFBs). Spontaneous activity was measured for 4 seconds using the macroscopic recording system (Fig. 3). Electrical impulses in control samples originated from the periphery of the monolayer and propagated uniformly with θ = 21cm/s (Fig. 3A). After the addition of 0.5 mmol/L of TC for 10 minutes (Fig. 3B), θ was significantly reduced to ∼9 cm/s and propagated APs displayed early afterdepolarization (EAD). In the particular experiment shown, the establishment of sustained re-entrant activity, displaying a rotation frequency of 5.2 Hz (see Supporting Video 1), followed these EADs.
UDCA Rescues TC-Induced Slow Conduction.
We have previously demonstrated that 0.1 mmol/L of UDCA resynchronizes arrhythmic cardiac calcium dynamics induced by 1 mmol/L of TC.18 We investigated the effect of UDCA alone and in combination with TC on impulse-propagation characteristics in the MH and FH models. Figure 4A shows that chronic exposure to either 0.1 or 1 μmol/L of UDCA had no effect on θ in MH models. However, in the FH model, UDCA, at concentrations from 0.1 to 100 μmol/L, increased θ significantly, as compared to untreated strands (Fig. 4B). Interestingly, the lowest concentration of UDCA (0.1 μmol/L) maximally improved θ. We next studied whether UDCA not only would improve θ, but would also protect the FH model from arrhythmias induced by either acute or chronic exposure to pathologically relevant concentrations of TC (0.5 mmol/L) (Fig. 4C,D). In the FH model exposed to TC either acutely or chronically, 0.1 μmol/L of UDCA mitigated the reduction of θ and, as a result, protected against the development of arrhythmias (see Supporting Video 2). After acute or chronic administration of TC plus UDCA, θ remained virtually unchanged (Fig. 4C,D). Interestingly, in preparations that underwent chronic pretreatment with UDCA, but from which UDCA was withdrawn before the addition of TC, protection against the slow conduction induced by TC was lost (Fig. 4D). This indicates that the continuous presence of UDCA, and not the treatment itself, is a prerequisite for cardioprotection. Table 1 shows a summary of the changes in impulse-conduction characteristics in the MH and FH models. Under acute conditions, increasing TC from 0.1 to 0.5 mmol/L led to (1) a decay of θ by up to 40.6%, (2) a decay of dV/dtmax by up to 19.3%, and (3) a decay of Vm by up to 28.1%. During chronic administration of TC, these parameters were further reduced, which corroborates the toxic effect of constant exposure to TC. In contrast, the presence of 0.1 μmol/L of UDCA during both acute and chronic administration of TC produced a recovery of θ, dV/dtmax, and Vm.
Table 1. Effect of Bile Acids on Impulse Conduction Characteristics
Conduction Velocity, cm/s
Upstroke Velocity, %APA/ms
Membrane Potential, mV
Measured in confluent srands of myofibroblasts-coated cardiomyocytes (n=7 each)
To understand the mechanisms of UDCA-induced cardioprotection in the FH model, we measured resting Vm using SICM in (1) the MH model, (2) the FH model, and (3) in MFB monolayers where the contamination by CMs was <1%, as shown by double immunostaining for desmin (green; CMs) and vimentin (red; fibroblastic cells; Fig. 5A). Although microelectrode impalements in MH and FH models were straightforward because of the models' substantial thickness, this was not the case for the flat MFB monolayers. To facilitate impalement, we first used hopping mode SICM17 and scanned a membrane region of 80 × 80 μm to image the surface topography of the MFB monolayers (Fig. 5B). MFBs were mostly flat when forming a monolayer, but some regions, presumably located above nuclei, were raised. We chose the highest point on a cell and used the same nanopipette for the impalement and measurement of Vm.
We found that acute exposure to 0.1 μmol/L of UDCA did not significantly affect Vm in the MH and FH models. However, in pure MFB monolayers, Vm significantly hyperpolarized from −28.2 to −41.9 mV (Fig. 5C). Notably, chronic exposure to 0.1 μmol/L of UDCA did not affect Vm in the FH model, but significantly hyperpolarized the FH model from −54.3 to −64.6 mV and MFB monolayers from −25.5 to −44.7 mV (Figs. 5D and 6A). We also demonstrate that glibenclamide (5 μmol/L), which binds to the sulfonylurea receptors (SURs), completely abolished the effect of UDCA on the Vm of MFBs (Fig. 6A). This suggests that UDCA-induced cardioprotection was dependent on a MFB-specific increase in potassium (K) conductance. In our view, adenosine triphosphate–gated K+ (KATP) channels are the most likely candidates, as sulfonylurea, at the tested concentrations, have been shown to be specific for KATP channels. Other currents present in the heart, such as inward-rectifying background K+ current (IK1) or inward L-type calcium current (ICa), are affected by a concentration of sulfonylurea nearly 10 times higher than the one shown to reduce IK(ATP) by 50%.19 To directly address this hypothesis, we performed radioligand-binding experiments with [3H]-glibenclamide and MFB on cell membranes, and demonstrated that UDCA can actively displace the radioligand from the SUR with an apparent half-maximal displacement at ∼0.1 μmol/L (Fig. 6B). This suggests that UDCA can directly bind to SUR, known to be overexpressed in cardiac MFBs,20 thereby regulating potassium conductance.
Our results obtained in the MH and FH models strongly suggest that MFBs play a pivotal role in the development of FAs during ICP and may provide an interesting, novel therapeutic target. The results also show that UDCA is a potential antiarrhythmic agent for treating FA (and, possibly, for arrhythmia in adult hearts). In particular, this study shows that (1) during gestation, MFBs transiently appear in the heart, (2) TC, at concentrations comparable to those found in ICP, provokes arrhythmogenesis and establishes re-entrant activity, and (3) UDCA protects against ICP-induced arrhythmia by directly hyperpolarizing MFBs.
MFBs in Fetal Human Hearts.
MFBs, also described as a “scarring phenotype,” are cells present in fibrotic10 and infarcted regions of the heart.21 They are likely to contribute to arrhythmogenesis by inducing nonuniform, discontinuous conduction after excess deposition of extracellular matrix. Furthermore, they affected the Vm of coupled CMs in vitro, slowing conduction and reducing ectopic activity.13, 14, 22 Oxygen levels play a key role for fibroblast transdifferentiation. Although mouse adult cardiac fibroblasts cultured at 21% O2 expressed de novo α-SMA,23 another study found no expression of α-SMA in human fetal cardiac and lung fibroblasts under the same conditions.9 We found that irregular α-SMA-positive regions appear in the human FH at 9 weeks (1.31%), becoming more abundant at the second and the third trimesters when ICP occurs (17.1%). Note that α-SMA-positive regions also stain positive for vimentin and DDR2, indicating that they are populated by noncardiac cells (i.e., MFBs). This is of relevance to ICP, as the disorder typically presents in the third trimester. Moreover, we previously demonstrated in vitro that MFB-induced arrhythmogenic ectopic activity occurs when the density of MFBs is >15.7%,14 corresponding, according to immunohistochemistry results, to the tissue abundance of MFBs at 15-26 weeks of gestation. Although previous studies indicate that α-SMA is transiently expressed during the early stage of the fetal life,24, 25 another study26 has shown no α-SMA positivity and fibrosis in septal and ventricular FH at 22 weeks of gestation. Nonetheless, the fetal MFB phenotype is transient, as MFBs are found neither in postnatal nor adult hearts. This transition may vary from heart to heart and can relate not only to the state of hypoxia, but also to MFB apoptosis, known to be the key to MFB turnover in cardiac valve development.27 This is in contrast to infarcted hearts, where the MFB phenotype persists for years.28 The abundance of MFBs in FH tissue at later stages of gestation and the defined border zone denote an intimate contact between the two cell populations, which could permit the establishment of heterocellular gap junctions. However, as is the case for MFBs in infarcted hearts, this has not been directly demonstrated in vivo, so far.
MFBs and Arrhythmia in ICP.
FA is found in as many as 2% of all pregnancies.29 Less than 10% of fetuses are found to have sustained tachyarrhythmias or bradyarrhythmias.30 The proposed etiology of FAs is heterogeneous and includes maternal metabolic disease,31, 32 structural abnormalities,33 and autoimmune disorders.34 It has previously been shown that MFBs tend to depolarize CMs after the establishment of heterocellular gap junctional coupling that, as a consequence, affect θ12, 13 and induce spontaneous ectopic activity.14, 21 However, it has not been previously addressed whether MFB might be involved in ICP-related FAs. We use a previously described in vitro model14 consisting of (1) monolayer strands of CMs (a simplified model of the adult MH devoid of MFBs) or (2) monolayer strands of CMs covered with a layer of cardiac MFBs (a simplified model of FH in the third trimester of gestation). Experimental conditions (i.e., CM isolation and seeding and superfusion solutions) were similar to that previously reported.12–14, 35 In the present study, we added TC to the MH model to generate an in vitro model of ICP that mimics the increase of TC in the bloodstream of pregnant mothers.36 In addition to pruritus and chronic hepatitis, such an increase may also cause cardiac dysrhythmia37 and unexplained stillbirth. Recently, we have described that TC induces arrhythmia in cultured single isolated CMs by targeting calcium dynamics via the muscarinic M2 receptor.7 The data presented here suggest that this mechanism of TC-induced arrhythmia does not apply to the MH model consisting of strands of confluent CMs. This is likely related to the fact that CMs coupled electrotonically to neighboring CMs in a confluent monolayer behave differently to single isolated CMs.14 Whereas single unloaded CMs can be expected to fire APs whenever Vm is disturbed to any significant degree (e.g., Ca2+ surge under Ca2+ overload conditions), similar excitations are suppressed in an electrotonically coupled network of CMs, as individual cells are clamped to a common resting potential. In the FH model, however, TC aggravates slow conduction in a concentration-dependent manner by, probably, affecting the calcium-based propagation, which is a major determinant of impulse conduction in these partially depolarized preparations.13 TC-induced slowing of θ and dV/dtmax was accompanied by a further depolarization (Table 1). This slowing of θ, together with the precipitation of EADs, ultimately gave rise to re-entrant electrical activity. The effects of TC on the electrophysiological parameters of the preparations were more pronounced during chronic than during acute administration and revealed a threshold for adverse effects on θ, dV/dtmax, and Vm at TC >0.3 mmol/L.
Cardioprotective Role of UDCA.
UDCA treats pruritus and hepatic dysfunction accompanying ICP.38 Our study shows, for the first time, that the effect of UDCA on impulse propagation characteristics depends upon the presence of MFBs (Fig. 5). Chronic exposure of the FH model to UDCA, ranging from 0.1 to 100 μmol/L, enhanced θ by roughly 2-fold. Acutely administered, UDCA partially counteracted the effects of TC on θ and dV/dtmax. In contrast, chronically administered UDCA completely abolished the effect of 0.5 mmol/L of TC, indicating a long-term effect on ion channels involved in conduction. Pretreatment of the preparations with UDCA before TC addition could not elicit this rescuing effect, but it was dependent upon the simultaneous presence of UDCA and TC. Regarding the mechanism underlying the observed UDCA effect, we found that neither acute nor chronic incubation with UDCA changed the diastolic Vm in the MH model, whereas it produced a substantial hyperpolarization in the FH model. This suggests that UDCA directly targets MFBs. MFBs are cells with a low resting Vm (see Figs. 5 and 6A),13 which is subject to modulation by axial stretch,39 hypoxia,40 and humoral activity.41 Our data show that UDCA is another modulator of MFB Vm, causing hyperpolarization both during acute and chronic application. One possible mechanism underlying hyperpolarization relates to an increase in potassium conductance by direct binding of UDCA to the SUR expressed in cardiac MFBs20 and is associated with various inwardly rectifying potassium ion channel (Kir) and KATP subunits. Note that in MFB-coated CM strands, the opening of KATP channels increases the degree of polarization and, invariably, stops arrhythmogenic spontaneous activity.14 In our study, even if KATP and Kir subunits were known to be present in CMs, UDCA exposure did not further hyperpolarize the well-polarized CM monolayer. Nevertheless, our finding of a hyperpolarization of an MFB monolayer after exposure to UDCA suggests that such a cellular compartment is affected and hence contributes to the overall antiarrhythmic effect of UDCA. Apart from a few studies, demonstrating that some aspects of fetal outcome may be improved in ICP patients receiving UDCA, compared to placebo or cholestyramine therapy,42, 43 the safety and the fetal benefit has not been adequately tested. A pilot study for a trial regarding the fetal benefit of UDCA is currently ongoing, recruiting cholestatic women in six centers.44 The fetus is recognized as a patient within a patient; therefore, every pharmacological treatment targeting FA may affect also the healthy MH. Our findings demonstrate that UDCA can protect indirectly against ICP-induced arrhythmogenesis by affecting MFBs, which are transiently present only in the FH. Thus, the findings of the present study might be significant for the clinical use of UDCA, as they reveal new mechanisms and a possible new strategy for the prevention of fetal cardiac arrhythmia.
Although the presence of MFBs in the ventricles of the human FH in the second and the third trimester is clearly demonstrated by the expression of α-SMA, further studies are required to characterize their appearance throughout gestation. The scarcity of human fetal cardiac tissue at later gestational weeks limits this approach. Moreover, it is still unclear which role MFBs might play during structural development of the heart and at which stage of gestation they disappear, as they are not present in newborn human hearts, except for the cardiac valve leaflets. Of particular interest and similar to our study, MFB density was found to be higher in cardiac valve development in the second trimester, accompanied by a progressive decrease via apoptosis with advancing gestation.27
Therefore, the in vitro model used in this study could only approximate the in vivo pathophysiological situation. In particular, it is not yet known whether MFBs in vivo establish heterocellular electrotonic coupling with CMs. Nevertheless, it has been found that fibroblasts in an infarct scar (known to be MFBs; cf. Fig. 1C and References21) express Cx43 and Cx45,45 and that MFBs in tissues different from the heart express connexin.46 These findings suggest, but do not prove, that heterocellular electrotonic coupling is present in the heart. Another limitation of our in vitro model of the FH is that it does not account for the specific 3D architecture of regions where MFBs intermingle with CMs in situ. Apart from technical difficulties related to the engineering of such structures, there is also a lack of information regarding the exact cytoarchitecture of these regions of the heart.
In conclusion, the results of the present study raise the possibility that the transient appearance of MFBs in the FH may increase the likelihood of FA during ICP. In general, fetal cardiac arrhythmia is treated with conventional antiarrhythmic drugs, which may have an adverse effect on the MH. Accordingly, UDCA may represent a new, specific therapeutic role in the management of FA. Indeed, UDCA therapy for ICP, aimed at improving serum bile acid levels, improves some aspects of fetal outcome. It remains to be investigated whether the benefit on fetal cardiac electrophysiological parameters mediated by MFBs might also be involved in this benefit to the fetus.
We thank Christian Dees for radioligand-binding assays and Regula Flueckiger-Labrada for her expert technical assistance. We also acknowledge the encouragement and enthusiasm by the late Professor Philip Poole-Wilson.