Introduction

Top Abstract Introduction Materials and Methods Results Discussion References

Considerable evidence has accrued that a variety of noncardiac drugs may prolong the QT interval of the surface electrocardiogram, which represents ventricular depolarization and repolarization, imparting an increased risk of developing a potentially fatal cardiac arrhythmia known as torades de pointes (TdP) (De Ponti et al., 2001). This has stimulated intense discussions and focused attention on methodological issues involved in the cardiac risk-benefit assessment of noncardiac pharmaceuticals, including antimicrobial agents, which has resulted in withdrawals of approved antibiotics such as grepafloxacin. So far, numerous electrophysiological studies have been conducted with respect to individual drugs to delineate the cellular basis of this electrophysiological phenomenon. Interestingly, although in theory the prolongation of action potential can result from disturbance of any of the cardiac ion channels, almost all reported QT-prolonging drugs that have been tested so far appear to inhibit I_Kr, the rapidly activating delayed rectifier current (De Ponti et al., 2001). This highlights the importance and uniqueness of I_Kr channel in cardiac repolarization. In humans, I_Kr is most likely carried by the potassium channel encoded by HERG (Sanguinetti et al., 1995). Drugs associated with QT prolongation and, occasionally, TdP have been reported to significantly inhibit the HERG-encoded channel current (Crumb and Cavero, 1999; Cavero et al., 2000). Due to inherent difficulties associated with I_Kr recording in native cardiac myocytes, mammalian cell lines expressing HERG have been widely used to assess the potency of drugs in inhibiting this channel and the cardiac electrophysiological safety of pharmaceuticals.

Macrolides are a group of closely related compounds characterized by a macrocyclic lactone ring (usually containing 14 to 16 atoms) to which sugars are attached. They have been widely used as effective antibiotics against Gram-positive organisms. However, numerous reports have been published regarding the cardiac adverse effects of macrolides, especially erythromycin (Katapadi et al., 1997). In fact, this prototype macrolide is also the most thoroughly characterized with respect to the effects on cardiac repolarization that seems to be mediated by cardiac potassium channel blockade (Antzelevitch et al., 1996; Rampe and Murawsky, 1997; Drici et al., 1998). In contrast, little is known about the effects of these macrolides on the HERG channel. With increasing reports of QT prolongation and arrhythmias associated with these drugs (Sekkarie, 1997; Lee et al., 1998; Kamochi et al., 1999; Woywodt et al., 2000), it is of particular importance to elucidate their electrophysiological characteristics. Therefore, we examined the effects of six macrolides on HERG currents expressed in HEK-293 cells and used clarithromycin to better understand the mechanism and relative role of HERG blockade by these macrolides.

	Materials and Methods

Top Abstract Introduction Materials and Methods Results Discussion References

Cell Preparation and Chemicals. HEK-293 cells stably expressing HERG potassium channels (Zhou et al., 1998) (licensed from Wisconsin Alumni Research Foundation) were cultured in minimal essential medium (Invitrogen, Carlsbad, CA) supplemented with 0.1 mM nonessential amino acid (Invitrogen), 1.0 mM sodium pyruvate, 10% fetal bovine serum and 0.05% geneticin (G418). On the day of the experiment, cells were dissociated using 0.05% trypsin-EDTA and stored at room temperature in M199 medium (Hanks' salt) for electrophysiological study. Erythromycin was purchased from Sigma-Aldrich (St. Louis, MO). All other drugs were provided by Pfizer Global Research and Development (Groton, CT). Drugs were dissolved in dimethyl sulfoxide at 10 to 30 mM as a stock solution, and then directly added into Tyrode's solution to a desired concentration. Dimethyl sulfoxide at the maximal concentration in this study (0.3%) did not have any detectable effect on the HERG current.

Patch-Clamp Recording. Aliquots of cells were allowed to settle on the bottom of the recording chamber (<0.5 ml in volume) mounted on an inverted microscope (Axiovert S100; Carl Zeiss, Inc., Thornwood, NY). Cells were superfused with Tyrode's solution containing 137 mM NaCl, 4 mM KCl, 1.8 mM CaCl₂, 1 mM MgCl₂, 10 mM glucose, and 10 mM HEPES (pH 7.4 with NaOH). Ionic currents were measured in the whole-cell configuration using a patch-clamp technique (Hamill et al., 1981). Recording electrodes with a resistance of 2 to 3 M when filled with the internal solution were connected to an EPC-9 patch-clamp amplifier controlled by the Pulse + PulseFit program (HEKA Elektronik, Lambrecht/Pfalz, Germany). The internal solution was composed of 130 mM KCl, 5 mM MgATP, 1.0 mM MgCl₂, 10 mM HEPES, and 5 mM EGTA (pH 7.2 with KOH). Seal resistances in all of the experiments were more than 1 G; therefore, leakage subtraction was not performed. Series resistance (R_s, generally between 3 and 6 M before compensation) was routinely compensated by at least 80% and checked periodically during the experiment. The anticipated voltage errors resulted from the uncompensated R_s in each experiment were limited to 5 mV. All experiments were performed at 35 ± 1°C. The bath temperature was maintained by a TC-344B temperature controller (Warner Instruments, Hamden, CT).

	Results

Top Abstract Introduction Materials and Methods Results Discussion References

Inhibitory Effect of Macrolides. The effects of six marketed macrolides and a metabolite of erythromycin, desmethyl erythromycin, on the HERG current are illustrated in Fig. 1, and typical recordings obtained with clarithromycin are shown in Fig. 1, A and B. The HERG current blockade by these macrolides was investigated in four to six cells at each drug concentration using the voltage protocol shown in Fig. 1A. Drugs were administered in a cumulative fashion up to a maximal concentration of 100 µM to eliminate the potential concerns of drug solubility. Concentration-response relationships, shown in Fig. 1D, were obtained by fitting the data with a Hill equation to yield IC₅₀ values. The resulting data are as follows (Hill coefficient in parentheses): clarithromycin, 32.9 µM (0.98); erythromycin, 72.2 µM (1.04); roxithromycin, 36.5 µM (1.16); josamycin, 102.4 µM (1.14); erythromycylamine, 273.9 µM (0.96); oleandomycin, 339.6 µM (1.13); and DME, 147.1 µM (0.88). The potencies of these compounds can therefore be rank-ordered as follows: clarithromycin  roxithromycin > erythromycin > josamycin >DME > erythromycylamine > oleandomycin. A Hill coefficient close to 1.0 was obtained for each compound, indicating a single binding site to the HERG channel.

View larger version (21K): [in this window] [in a new window]   Fig. 1.   Effects of macrolides on the HERG current. A, superimposed current traces in an experiment with clarithromycin. Currents were elicited by a 1-s depolarization step to +20 mV from a holding potential of 80 mV, followed by a repolarizing ramp (0.5 V/s) to 80 mV. Stimulation frequency, 0.25 Hz. B, time course of the effect of clarithromycin. A linear regression was performed through the linear portion of control values and extrapolated over the entire experiment. Dofetilide (10 µM) was applied at the end of the experiment to evaluate the endogenous outward currents, which were subtracted off-line. The numbers in panels A and B indicate the absence (0) and presence of clarithromycin at 3, 10, 30, and 100 µM (1, 2, 3, and 4, respectively), and the addition of dofetilide (5). C, concentration-response relationships of macrolides on HERG. Steady-state blocks of the macrolides were plotted against testing concentrations. Data were fitted using the Hill equation. CLM, clarithromycin; ERM, erythromycin, JSM, josamycin; EMA, erythromycylamine; OLM, oleandomycin; RXM, roxithromycin.

Voltage- and Time-Dependent Block by Clarithromycin. Given the similarity in the chemical structure to other macrolides and the increasing attention to its cardiac side effects, clarithromycin was chosen for further mechanistic exploration. Figure 2A shows representative current recordings in response to a series of depolarization steps before and after the administration of 30 µM clarithromycin. Cells were held at 80 mV and stimulated with a series of 1-s depolarizing pulses ranging from 70 to +60 mV with 10-mV increments at 0.1 Hz. Tail currents were recorded upon repolarization to 60 mV. When measured at the end of the depolarizing steps, the time-dependent current started to activate at voltages between 60 and 50 mV, peaked at 20 mV, and then showed inward rectification because of the rapid voltage-dependent C-type inactivation (Smith et al., 1996). Addition of 30 µM clarithromycin significantly inhibited the HERG current at voltages positive to 30 mV. Current reductions at 30, 0, and 40 mV were 18 ± 10, 45 ± 5, and 56 ± 7%, respectively (n = 12; Fig. 2C). In 8 of 12 cells studied, increases in the steady-state current amplitude of 97 ± 31 and 36 ± 15%, respectively, were observed at 50 and 40 mV, which accounted for the elevation and the large variation of the total averaged currents at these two voltages. This phenomenon could not be explained by experimental artifacts, such as changes in pipette-membrane sealing conditions and series resistance. Moreover, the increase in amplitude was only observed at voltages close to the threshold of channel activation (50 and 40 mV). In contrast, when a series of repetitive depolarization steps to 50 mV were applied in the presence of the drug, clarithromycin only caused a reduction of the current amplitude (data not shown). These observations are similar to those described previously for azimilide (Jiang et al., 1999), which may reflect two separate drug actions on the HERG channel through different sites. To estimate the potential contribution of the enhancing effect of clarithromycin to the pattern of the "voltage-dependent" reduction of the current amplitude, averaged data from the other four observations were plotted in Fig. 2C (open symbols). Increased inhibition was again observed as the membrane was depolarized (from 25% inhibition at 50 mV to 65% at +60 mV, respectively).

View larger version (37K): [in this window] [in a new window]   Fig. 2.   Effect of clarithromycin 30 µM on the steady-state and tail HERG current at various membrane potentials. A, representative current recordings in response to a series of voltage steps (upper panel) in the absence (open circle) and presence (closed circle) of clarithromycin (30 µM). B, effect of clarithromycin on the steady-state current-voltage (I-V) relationship of HERG current. Current amplitudes at the end of 1-s depolarization steps before () and after () clarithromycin (30 µM) were measured, averaged (n = 12), and plotted against membrane potentials. C, voltage-dependent blocking of the HERG current by clarithromycin (30 µM). Ratios of the current amplitudes after clarithromycin to the controls at each membrane potential were obtained from a total of 12 observations (), among which eight cells showed a small potentiation of HERG current at 50 and 40 mV. Results from the remaining 4 experiments were shown as the open squares (). D, effect of clarithromycin on the voltage dependence of steady-state activation of HERG. Peak tail currents at each voltage were measured and normalized to the maximal currents. Data were fitted with a single Boltzmann function which resulted in a V1/2 and an s (slope factor) of 33.4 ± 0.9 and 5.4 ± 0.1 mV in control, and 38.2 ± 0.8 and 5.7 ± 0.1 mV in the presence of clarithromycin (n = 12, p < 0.01)

View larger version (24K): [in this window] [in a new window]   Fig. 3.   Clarithromycin block of HERG does not require channel activation. A and B, and C and D, respectively, show the original recordings and the time courses of E-4031 and clarithromycin effects at concentrations close to their IC50 values. Cells were held at 80 mV for 3 min to keep the channel at closed conformation during the initial superfusion with E-4031 or clarithromycin.

View larger version (17K): [in this window] [in a new window]   Fig. 4.   Clarithromycin does not affect the kinetics of the HERG channel. Time constants of HERG channel activation (act), deactivation (deact), inactivation (inact), and recovery from inactivation (rcvy) before (open symbols) and after (closed symbols) administration of clarithromycin were obtained. No significant difference was found in any of the parameters. Data were obtained from four to seven observations.

Kinetics of Block by Clarithromycin. The blockade of the HERG current by clarithromycin, as shown in Figs. 2A and 5A, appeared to develop gradually during the depolarization steps, indicating that the inhibitory effect of clarithromycin is time-dependent. To further investigate the blocking time course, currents recorded in the presence of the drug were normalized to the control currents, and the resulting ratio of relative current amplitudes at each membrane potential was plotted (Fig. 5B). In this figure, the resulting relative currents from a representative cell are shown following depolarizing steps to 40, 20, +20, and +50 mV, respectively. These data could be well described by a single-exponential function (Fig. 5B, solid line), and the resulting time constants and the steady-state relative amplitudes at different potentials are summarized in Fig. 6, A and B, respectively. The results again show that the inhibition of HERG current by clarithromycin is both voltage- and time-dependent.

View larger version (20K): [in this window] [in a new window]   Fig. 5.   Voltage dependence of the time courses of clarithromycin-induced inhibition of HERG current. A, superimposed current traces during depolarization to 40, 20, and +20 mV in the absence and presence of 30 µM clarithromycin. B, time courses of clarithromycin-induced inhibition. Current ratios of IDrug to ICtrl were obtained at each membrane potential and fitted with a single-exponential function (solid lines).

View larger version (21K): [in this window] [in a new window]   Fig. 6.   Analysis of clarithromycin block of HERG channel currents. A and B, time constants () and the steady-state relative current at the end of the depolarization step (Po) obtained from exponential fittings as described in the legend to Fig. 6. C and D, unbinding () and binding () rate constants, which were calculated from Po and : Po = /( + ),  = 1/( + ). Both  and  were fitted with a Boltzmann equation, resulting in a half-maximal voltage (V1/2) of 13.1 mV and a slope factor (s) of 5.2 mV for , and V1/2 of 9.0 and s of 9.9 mV for .

	Discussion

Top Abstract Introduction Materials and Methods Results Discussion References

The pharmacokinetic parameters of the macrolides tested in this study following oral administration at relatively high doses, as illustrated in Table 1, are compared with their potencies (IC₅₀) to inhibit the HERG channel. Calculated ratios of IC₅₀ to C_max for individual drugs range from 15- to 642-fold. However, inhibition of I_Kr and QT prolongation by these drugs may still occur at clinically relevant concentrations for the following reasons. First, QT prolongation has been reportedly induced by drugs at concentrations causing <20% inhibition of HERG. Cisapride, for instance, produced an average of 6 ms of QTc prolongation over 24 h in healthy volunteers at 40 ng/ml (van Haarst et al., 1998). At this concentration (free drug, 1.7 nM), it only inhibited the HERG current by 10 to 20% (Mohammad et al., 1997; Rampe et al., 1997). Several other drugs, including E-4031, dofetilide, terfenadine, and risperidone, are also reported to produce QT prolongation at clinical concentrations significantly lower than the IC₅₀ values reported to inhibit the HERG current. Second, clinical exposure of macrolides can often exceed the concentrations listed in Table 1, in cases such as intravenous administration, DDIs, and poor metabolism (Rubart et al., 1993; Sekkarie, 1997; van Haarst et al., 1998). The peak serum concentrations of erythromycin, for example, can average ~30 µg/ml after a 900-mg i.v. infusion compared with 2 to 4 µg/ml after a 500-mg oral dose (Rubart et al., 1993). Finally, most of the macrolides studied here are found to accumulate in tissues, including the heart, and the resulting local concentrations can be significantly higher than in the plasma (Yoshida and Furuta, 1999). Nonetheless, clarithromycin, roxithromycin, and erythromycin (IC₅₀ values were 50-fold C_max) would then be expected to have a higher propensity than the other three drugs to cause QT prolongation in humans according to the rank order of potency. In fact, they all have been associated with long QT syndrome and arrhythmias, whereas no adverse cardiac events have yet been reported for josamycin, oleandomycin, and erythromycylamine. Therefore, although inhibition of HERG current appears to be a common feature for these macrolide antibiotics, differences in their potencies and therapeutic windows seem to be the key to the clinical outcomes.

View this table: [in this window] [in a new window]   TABLE 1 Pharmacokinetic parameters and potencies of HERG inhibition of the macrolides Data were obtained from Periti et al. (1989), Nilsen (1995), and McFarland et al. (1997). Data were also obtained from Physicians' Desk Reference (1999).

The effect of a drug on QT interval can be influenced by various factors, such as multiple ion channel activity, electrolyte disturbances (e.g., hypokalemia or hypomagnesemia), bradycardia, genetic defects underlying the congenital long QT syndrome, female gender, and concomitant medications that may prolong the QT interval and/or affect the pharmacokinetics of the drug. Extrapolation of the in vitro study on a single ionic current, therefore, has to be made cautiously. In this study, we observed that a metabolite of erythromycin, DME, also blocked the HERG channel with an IC₅₀ of 147.1 µM, implying that only comparing the potencies among parent drugs may be an oversimplification. Importantly, all these macrolides can inhibit and/or be metabolized by liver enzymes, especially cytochrome P450 subtype CYP3A4, with a tendency to cause DDIs in the following order: erythromycin > clarithromycin > josamycin  roxithromycin > erythromycylamine  oleandomycin (Periti et al., 1992). DDIs resulting in relevant cardiac events have indeed been reported between macrolides and multiple agents (Pai et al., 2000; Westphal, 2000). Moreover, mutant forms of MinK-related peptide 1, an accessory peptide likely coassembling with the pore-forming HERG subunit to form the native I_Kr (Abbott et al., 1999), has reportedly demonstrated diminished potassium currents and/or increased channel blockade by clarithromycin (Abbott et al., 1999; Sesti et al., 2000). Therefore, some instances of antibiotic-associated QT prolongation and TdP may in fact represent the unmasking of otherwise silent genetic channelopathies. Furthermore, lack of clinical information resulting from limited utilization of a drug and insufficient awareness among physicians of prescribed QT-prolonging drugs (Yap and Camm, 2000) may also make the correlation difficult. Among macrolides, clarithromycin was listed with the highest report-utilization ratio of 0.34 cardiac event reports per million prescriptions by the Adverse Event Report System of the U.S. Food and Drug Administration, whereas erythromycin accounts for 53% of the total reports among macrolides because of its long cumulative time on market. Roxithromycin, which is not registered in the U.S. market, has only a few clinical reports, so far, on QT-related adverse effects (Ortqvist et al., 1996; Woywodt et al., 2000), although it is a relatively potent HERG blocker, as shown in this study. All things considered, our data do not exclude the possibility of QT prolongation and cardiac arrhythmias being associated with josamycin, erythromycylamine, and oleandomycin.

The aromatic residues (e.g., Y652 and F656) of the S6 domain of the HERG channel have recently been demonstrated to interact hydrophobically with class III antiarrhythmic drugs such as dofetilide (Mitcheson et al., 2000a). In a competitive ligand-binding study, we observed that macrolide compounds, including erythromycin, showed a concentration-dependent inhibition of [³H]dofetilide binding to the HERG channel in micromolar ranges, indicating that they may interact with the same site(s) as dofetilide (data not shown). Interestingly, erythromycin, clarithromycin, and roxithromycin have higher lipophilicity (E log Ds are 1.3 to 1.6 at physiological pH) than the other macrolides (0.31 and 0.43 for erythromycylamine and oleandomycin, respectively) (McFarland et al., 1997), which may favor interactions with the binding site and correlate with their relatively higher potencies. We also showed that clarithromycin block of the HERG channel was slightly voltage-dependent, a property commonly seen for the classical I_Kr blockers such as E-4031. These blockers require activation of the channel to access the inner binding site and get "trapped" when the channel is closed (Mitcheson et al., 2000a,b). Therefore, the kinetics of both blocking and unblocking are normally very slow (>10 min to reach a steady state in our experiment). However, clarithromycin, despite its large molecular size, can inhibit the current even when channels are closed and the recovery from blocking is fairly fast (<3 min washout in this study), indicating that it may be able to bypass the activation gate or the "trapping" mechanism through a hydrophobic pathway. Certainly, more experiments may be needed to reach a definitive conclusion on its interaction with the channel. Since clarithromycin has a pK_a of 8.99, it may be positively charged at physiological pH and, therefore, may have an electrostatic response to the depolarized voltages. This may account for, at least in part, the voltage dependence of HERG channel blockade. Our kinetic analysis on clarithromycin block demonstrated a first-order reaction between the channel and drug molecules. This reaction is apparently independent of channel gating because a single-exponential fit is sufficient to describe the time course of the effect during depolarization steps. Besides, the difference between the voltage dependence of the binding rate constant (V_1/2 of 13.1 and s of 5.2 mV) and that of the HERG activation (V_1/2 of 33.4 and s of 5.7 mV) also indicates that channel blocking and activation gating are likely two separate processes. Because both binding and unbinding rate constants increase as membrane potential is more depolarized, it is likely that the drug-channel interaction may be affected by electrostatic responses in both ways (facilitating or hindering). We also observed that clarithromycin exhibited a small enhancing effect on the HERG current at voltages close to its activation threshold (50 to 40 mV), an effect similar to that of azimilide, reported previously (Jiang et al., 1999). Further studies are warranted to investigate its mechanism and implications.