State-dependent block of HERG potassium channels by R-roscovitine: implications for cancer therapy

Sindura B. Ganapathi, Mark Kester, Keith S. Elmslie


Human ether-a-go-go-related gene (HERG) potassium channel acts as a delayed rectifier in cardiac myocytes and is an important target for both pro- and antiarrhythmic drugs. Many drugs have been pulled from the market for unintended HERG block causing arrhythmias. Conversely, recent evidence has shown that HERG plays a role in cell proliferation and is overexpressed both in multiple tumor cell lines and in primary tumor cells, which makes HERG an attractive target for cancer treatment. Therefore, a drug that can block HERG but that does not induce cardiac arrhythmias would have great therapeutic potential. Roscovitine is a cyclin-dependent kinase (CDK) inhibitor that is in phase II clinical trials as an anticancer agent. In the present study we show that R-roscovitine blocks HERG potassium current (human embryonic kidney-293 cells stably expressing HERG) at clinically relevant concentrations. The block (IC50 = 27 μM) was rapid (τ = 20 ms) and reversible (τ = 25 ms) and increased with channel activation, which supports an open channel mechanism. Kinetic study of wild-type and inactivation mutant HERG channels supported block of activated channels by roscovitine with relatively little effect on either closed or inactivated channels. A HERG gating model reproduced all roscovitine effects. Our model of open channel block by roscovitine may offer an explanation of the lack of arrhythmias in clinical trials using roscovitine, which suggests the utility of a dual CDK/HERG channel block as an adjuvant cancer therapy.

  • S620T mutant
  • open channel block
  • terfenadine
  • human ether-a-go-go-related gene model

human ether-a-go-go-related gene (HERG) potassium channels (Kv11.1) are a critical regulator of the cardiac action potential (cAP) duration. HERG channels display unique kinetics that allows them to become fully activated during the repolarization phase of cAP (11). These channels activate slowly and inactivate rapidly during the initial phases of cAP to limit the outward potassium flux that would otherwise prematurely terminate the action potential (11). As repolarization begins, HERG channels rapidly recover from inactivation, and these reactivated channels generate the strong potassium efflux that completes the repolarization process (34). As a result, HERG channel block can prolong the cAP, causing acquired long QT syndrome that can lead to fatal torsades de pointes (34). HERG is blocked by a diverse group of chemical compounds, and many prescription drugs have been removed from market because of their HERG-blocking property (34). These drugs are thought to bind within a large inner cavity (vestibule) formed by the four S6 segments that can trap the drug when the channel closes (29). Trapping of drugs within the inner vestibule increases the block duration, which likely contributes to the proarrhythmic properties of these drugs. Therefore, a drug that is not trapped by HERG channels may have a reduced effect on cardiac rhythm.

In addition to regulating the cAP, HERG channels also play a role in cancer. Multiple hematopoietic cancer cell lines have been shown to regulate HERG, and blocking HERG reduces proliferation and tumor invasiveness (1, 24, 28, 41). HERG regulates proliferation in peripheral blood mononuclear cells and is constitutively expressed in primary human acute myeloid leukemias (30). In the human colon adenocarcinoma cell line, HT-29, the HERG blocker erythromycin selectively suppressed proliferation of cells expressing HERG channels, which resulted in synergistic cytotoxicity effects when erythromycin was combined with anticancer agents such as vincristine, paclitaxel, and hydroxy-camptothecin (8). Thus, HERG blockers have been proposed as an adjuvant cancer therapy (23, 31). In nonexcitable cells, the steady-state activation and inactivation vs. voltage relationships for HERG currents cross around −40 mV to reveal a voltage range that generates a steady-state current (window current), which is important in the maintenance of the resting membrane potential in nonexcitable cells and is thought to play a role in carcinogenesis (2). Together these findings support HERG as an attractive target for the antineoplastic therapy (1).

Roscovitine is a cyclin-dependent kinase (CDK) inhibitor that is currently undergoing phase II clinical trials as an anticancer drug (20, 35). We recently demonstrated that R-roscovitine (Rosc) blocks voltage-dependent potassium channels (Kv1.3, Kv2.1, and Kv4.3) by an open channel mechanism (6). We postulated that Rosc could also block HERG channels. Given that many diverse drugs block HERG channels by an open channel mechanism, we were not surprised to find that Rosc also blocked open HERG channels. However, unlike other drugs, Rosc was not trapped by HERG channel closing, which could explain the absence of arrhythmias in recent Rosc clinical trials (3, 18). Moreover, our studies support the use of Rosc as an adjuvant therapy for cancer due to the dual block of CDKs and HERG channels with a reduced proarrhythmogenic potential.


Cell culture.

Human embryonic kidney (HEK)-293 cells stably expressing HERG (a gift from Dr. Eckhard Ficker, MetroHealth Medical Center, Cleveland, OH) were maintained in DMEM + glutamax media (Invitrogen) supplemented with 10% FBS, 100× antibiotic-antimycotic, and the selection agent G418. Chinese hamster ovary (CHO) cells were used for transient transfection of S620T mutant (see below) and were maintained in F12 + glutamax media (Invitrogen) supplemented with 7.5% FBS and 100× antibiotic-antimycotic (without G418). For electrophysiology, cells were harvested using 0.25% trypsin/EDTA (Invitrogen) and plated into 35-mm culture dishes that served as the recording chamber (6).

Transient transfection.

The HERG mutant S620T cDNA clone used for transient transfection was very generously provided by Dr. Shetuan Zhang (Queen's University, Kingston, Ontario, Canada). Transient transfection of CHO cells utilized FuGENE 6 (Roche Applied Science) following the manufacturer's instructions. Mutant cDNA (0.9 μg) with 0.1 μg of green fluorescent protein (GFP) was combined with FuGENE 6 at a 6:1 reagent to cDNA ratio. GFP-expressing cells were selected for study.

Electrophysiological recording.

Membrane currents were recorded in whole cell patch-clamp mode at room temperature (∼23°C). Electrodes were pulled from Schott 8250 glass (inside diameter 0.90 mm, outside diameter 1.50 mm; Garner Glass, Claremont, CA) using a Flaming/Brown P-97 pipette puller (Sutter Instrument, San Rafael, CA). Electrodes typically had a resistance of 1–3 MΩ. Series resistance was compensated at ≥80%. Currents were amplified and filtered using an Axopatch 200A amplifier (Molecular Devices, Sunnyvale, CA) and digitized with either a 12-bit A/D converter (GW Instruments, Somerville, MA) or an ITC-18 (Instrutech, Bellmore, NY) after analog filtering with the amplifier's four-pole low-pass Bessel filter.

HERG current comprised >90% of the potassium current in the stably transfected HEK-293 cells since 0.1 μM terfenadine (Terf) blocked 91 ± 4% of the tail current at −40 mV (n = 22). The endogenous potassium current expressed by some HEK-293 cells was not affected by 0.1 μM Terf (+4 ± 11% effect on tail current at −40 mV, n = 9).


The internal solution contained (in mM) 120 KCl, 6 MgCl2, 10 N-methyl-d-glucamine (NMG)·HEPES, 5 NMG2·EGTA, 5 Tris2·ATP, and 0.3 Tris2·GTP. The pH was adjusted to 7.4 using NMG base and the osmolarity was 304 mosM. The external solution contained (in mM) 140 NaCl, 4 KCl, 2 CaCl2, 1 MgCl2, 10 NMG·HEPES, and 10 glucose. The pH was adjusted to 7.4 using NMG base, and the osmolarity was 313 mosM.

Data acquisition and analysis.

Data were taken using either S3 or S5 (software developed by Dr. Stephen Ikeda, National Institutes of Health, National Institute on Alcohol Abuse and Alcoholism, Bethesda, MD) on a Macintosh computer (Apple Computer, Cupertino, CA). IgorPro software (WaveMetrics, Lake Oswego, OR) was used to measure current amplitudes and to fit (Marquardt-Levenberg algorithm) equations to currents and data plots.

The step time course and trapping protocols were leak subtracted using scaled hyperpolarizing pulses of one-fourth amplitude (P/4). For all other protocols, leak current was estimated by linear regression of the nearest current-voltage (I-V) relationship over voltages (−80 to −40 mV) that did not generate HERG current. Data were not used if the holding current differed by >10% between the I-V used to generate leak and the protocol for which that leak was to be used for subtraction. Solutions were applied using a gravity-fed flow system with six inputs and a single output. The exchange time for this system is 1–2 s. For examining the time course of drug effects, cells were pulsed to +20 mV (1-s duration) once every 20 s, and currents were measured from peak tail currents at −40 mV following the +20-mV step (the step time course protocol). Running this protocol under control conditions revealed a slow time-dependent loss of HERG current in most cells (rundown), which appeared to result in the partial recovery of current amplitude on washout of roscovitine. If uncompensated, rundown would contribute to an overestimate of Rosc-induced inhibition. Since rundown cannot be independently measured during drug-induced inhibition, we estimated the effect of rundown by averaging current measurements (MControl) taken just before (MBefore) and on recovery from (MRecov) Rosc-induced inhibition (MDrug) [i.e., MControl = (MBefore + MRecov)/2]. We have previously found that using an average of measurements before and on recovery from drug application for MControl was best for compensating for the effect of rundown on fractional inhibition (14). Fractional inhibition was calculated from (MControl − MDrug)/MControl. MBefore, MDrug, and MRecov were each the average of three tail current measurements taken before, during, and on recovery from drug application, respectively.

Ratio currents provide a method to observe the change of inhibition with time during voltage steps (15). The ratio current was calculated from (IControlIDrug)/(IControl), where IControl is the average of IBefore and IRecov. IBefore, IDrug, and IRecov are each the average of three current traces before, during, and on recovery from drug application, respectively. Data are reported as means ± SD.

Computer simulations.

Simulated currents were generated using Axovacs 3 (written by Stephen W. Jones, Case Western Reserve University, Cleveland, OH) running on a Dell Inspiron 640m computer (Dell Computer, Round Rock, TX). Voltage (V)-dependent rate constants (kx) in the model were calculated from Math(1) where Ax is the rate constant at 0 mV, zx is the charge moved, and R, T, and F are the gas constant, absolute temperature, and Faraday's constant, respectively. The ionic concentrations for the current simulations were [K+]o = 5 and [K+]i = 140 mM.


R-roscovitine was purchased from LC Labs (Woburn, MA). Indirubin was obtained from EMD Biosciences. Terfenadine, NMG, KCl, MgCl2, HEPES, EGTA, Tris2·ATP, Tris2·GTP, NaCl, CaCl2, and glucose were purchased from Sigma (St. Louis, MO). Terf, Rosc, and indirubin were dissolved in DMSO to make a stock solution of 50 mM. For experiments using these drugs, all external solutions contained identical concentrations of DMSO.


Rosc inhibits HERG current.

HERG current block by many different drug classes is characterized by slow development (2–5 min) and a slow recovery (>10 min) (17, 40). This type of block is exemplified by the antihistamine Terf (32), which achieved steady-state block in 1–2 min and showed incomplete recovery even after a 9-min washout period (see Fig. 1). This is contrasted by Rosc, which blocked HERG current in ∼20 s and completely recovered ∼20 s following termination of drug application (Fig. 1). The speed of Rosc block suggested a unique mechanism relative to other blockers (e.g., Terf) and that kinase inhibition is not involved (5, 6). This hypothesis was further tested by examining the effect of the structurally unrelated CDK inhibitor indirubin-3′-monoxime (Indir), which has a similar affinity for CDK1 (0.2 vs. 0.5 μM) and CDK5 (0.1 vs. 0.2 μM) relative to Rosc (22). Indir (30 μM) showed a slower block and recovery compared with Rosc. Thus, Rosc appears to directly interact with HERG channels to block potassium flux.

Fig. 1.

Comparison of the effects of roscovitine, indirubin, and terfenadine on human ether-a-go-go-related gene (HERG). A: peak tail currents were measured from a representative cell (same as in B) showing the time course of inhibition and recovery from inhibition for 30 μM R-roscovitine (Rosc), 30 μM indirubin-3′-monoxime (Indir), and 0.1 μM terfenadine (Terf). On average, inhibitions were 49 ± 9% (SD) for Rosc (n = 10), 61 ± 11% for Indir (n = 4), and 89 ± 7% (n = 6) for Terf. Step currents were generated by 1-s steps to +20 mV, and tail currents were generated by a 1-s repolarization to −60 mV. The interval between sweeps was 20 s. Note the differences in the speed of inhibition induced in each of these drugs. B: representative current traces from A with the voltage protocol shown below. The arrowhead indicates the point at which peak tail currents were measured. Cntl, control; Recov, recovery.

The Rosc dose-response relationship was measured from the peak tail current (−60 mV) following a 1-s step to +20 mV (20-s interval) to activate HERG channels (Fig. 2). Rosc inhibited HERG current in a dose-dependent manner with little or no effect at 1 μM and almost complete block at 100 μM. The IC50 was calculated using the Hill equation to be 27 μM (n = 6–10) (Fig. 2B). This apparent affinity was similar to that for neuronal delayed rectifier potassium current (IC50 = 23 μM), but lower than that for expressed Kv2.1 (IC50 = 10 μM) (6). The Hill coefficient was 1.4, which indicates a small positive cooperativity in the Rosc block. The dose-response relationship measured from step current (at the end of the 1-s step) showed an IC50 = 35 μM and a Hill coefficient of 1.7. These higher values at a voltage where more HERG channels are inactivated could indicate that Rosc differentially affects inactivated channels (see below).

Fig. 2.

Concentration-dependent block of HERG by Rosc. A: a representative cell shows the time course of inhibition for 1, 3, 10, 30, and 100 μM Rosc. Step currents (○) were measured at the end of the 1-s step to +20 mV. The voltage protocol and peak tail current measurements (•) are the same as shown in Fig. 1B. B: dose-response relationship showing the fractional block of tail currents by Rosc. The smooth solid line is a fit using the Hill equation to generate IC50 = 27 μM and Hill coefficient = 1.4, while the dashed line shows the fit with Hill coefficient fixed to 1 (n = 6–10).

Rosc block depends on HERG channel activation.

The speed and dose-response relationship of block provide important clues regarding possible mechanisms. To continue investigating the blocking mechanism, we recorded the HERG I-V relationship by stepping the voltage from −60 to +40 mV (20-mV increments) and measuring current from both step (at 4 s) and tail currents (peak) (Fig. 3). The step current showed slow activation at negative voltages and a bell-shaped I-V relationship that is typical of HERG channels. This shape results from channels inactivating more completely at depolarized voltages to reduce current measured during the step (33, 36). Repolarization to −40 mV induced the characteristically large tail currents that result from the rapid recovery from inactivation. A plot of peak tail current vs. voltage showed that the threshold for current activation was −20 mV and that maximum activation was achieved at +20 mV (Fig. 3C). Rosc (30 μM) reversibly blocked HERG current, but the effect depended on the degree of depolarization (Fig. 3D). The block was weaker at −20 mV (mean = 30 ± 11%, n = 7) and reached a maximum at +20 mV (mean = 49 ± 5%, n = 7), which corresponded with maximal activation (Fig. 3). The block was associated with a small but significant left-shift in the voltage dependence of activation (shift in V1/2 = −2 ± 2 mV; P < 0.01, t-test paired samples, n = 10). These effects could be explained if Rosc block either was correlated with HERG channel activation or exhibited weak intrinsic voltage dependence.

Fig. 3.

Rosc block increases with step depolarization. A: HERG currents from a representative cell are shown for voltage steps to −20, 0, and +20 mV. The tail currents were generated on hyperpolarization to −40 mV. The Rosc concentration was 30 μM. Control (Cntl; before Rosc application), Rosc (during Rosc application), and recovery (Recov; on recovery from Rosc application) currents are represented. The arrowhead indicates the point at which peak tail currents were measured. B: the step current vs. step voltage relationship (I-V) is shown for the representative cell from A. Step currents were measured at the end of the 4-s step ranging from −60 to +40 mV (20-mV increments). Values are shown for control (▵), 30 μM Rosc (•), and on recovery (▿). C: peak tail currents are plotted vs. step voltage (symbols same as in B) for the same representative cell (A and B). The smooth lines represent single Boltzmann fits to generate V0.5 = −7, −11, and −7 mV, slope = 7, 7, and 7, and maximum current = 1.7, 0.8, and 1.5 nA for control, 30 μM Rosc, and recovery, respectively. D: comparison of the percent inhibition of peak tail currents induced by 30 μM Rosc shows significantly lower inhibition at −20 vs. 0 (b), +20 (c), and +40 mV (d) (P < 0.01 for each comparison, ANOVA; n = 6).

Interestingly, the voltage dependence of block was reversed when examining the effect of Rosc on tail currents measured at different voltages (Fig. 4A). HERG current was activated by a voltage step to +60 mV and tail currents measured at voltages ranging from +20 to −100 mV. The inhibition increased from 26% at +20 mV to 43% at −40 mV and 47% at −60 mV (P < 0.05 for +20 vs. −40 mV, and +20 vs. −60 mV, n = 5, ANOVA) (Fig. 4B). This combined with the result from the I-V protocol (discussed above) demonstrates that Rosc block is not voltage-dependent per se, but primarily depends on the availability of activated HERG channels.

Fig. 4.

Increased Rosc block of tail current with hyperpolarization. A: the peak tail current vs. tail voltage relationship is shown for a representative cell. Peak tail currents were measured at voltages ranging from 20 to −100 mV (20-mV increments) following a 1-s step to 60 mV to activate HERG channels (inset) and are plotted vs. tail voltage for control (▵), 30 μM Rosc (•), and recovery (▿). The arrowhead on the voltage protocol indicates the point at which the tail current was measured. B: percent inhibition of IHERG by 30 μM Rosc from the voltage protocol shown in A. Percent inhibition at different tail voltages is plotted vs. tail voltage. Inhibition at 20 mV was significantly different (*P < 0.05, n = 5) from that at −40 and −60 mV (ANOVA). C: Rosc decreases inactivation kinetics only at intermediate voltages. The recovery from inactivation τ (circles, n = 6) was measured (single exponential equation) from the raising phase of the tail current upon hyperpolarization from +60 mV to the indicated voltage (x-axis). The development of inactivation τ (squares, n = 4) was measured from a triple-pulse protocol where voltage was stepped to +60 (500 ms), −100 (10 ms), and the indicated voltage (x-axis) for 250 ms. A single exponential equation was used to fit inactivation during the third voltage step. Data are shown for control (average of values before and after Rosc) and during the application of 30 μM Rosc (open and closed symbols, respectively) (**P < 0.01).

Inactivated HERG channels are relatively insensitive to Rosc.

The increase in block with step voltage could result from Rosc block of inactivated HERG channels, which should be revealed by examining inactivation kinetics. Using the tail I-V protocol (described above), the recovery from inactivation time constant (τ) was determined by fitting the initial rising phase of tail current with a single exponential equation (33, 36, 47). Rosc increased the recovery from inactivation τ at voltages ranging from +20 mV to −40 mV, but not at −80 and −100 mV (Fig. 4C). The acceleration of recovery was observed only at voltages where τ was >6–8 ms, which could indicate that Rosc affects a process other than inactivation. The development of inactivation was measured using an inactivation protocol where voltage was stepped to +60 mV for 500 ms to inactivate channels, hyperpolarized to −100 mV for 5 ms to recover channels from inactivation, and returned to voltages ranging from −20 to +60 mV, which was where the inactivation τ was measured (45). Rosc (30 μM) significantly increased the speed of inactivation at voltages < +20 mV, but not at +20 to +60 mV where inactivation is extremely fast (Fig. 4C). Again, the differential effect of Rosc on inactivation kinetics may indicate that a separate process (e.g., open state block) is being affected.

Rapid Rosc application supports open state block.

To further investigate the possibility of open channel block, we used our rapid exchange flow system (6) to apply and remove Rosc during a single 10-s tail current at −40 mV (Fig. 5A). Rosc (30 μM) reversibly blocked the tail current by 38 ± 8% (n = 4). Thus, open HERG channels are blocked by Rosc. To compare the effect of Rosc on inactivated vs. open channels in the same cell, we used another protocol where HERG currents were elicited by an 8-s step to 60 mV to inactivate the channels followed by repolarization to −40 mV where majority of HERG channels recover from inactivation and enter the open state (Fig. 5B). Rosc (30 μM) was applied toward the end of the step to +60 mV and continued through the repolarization to −40 mV. Rosc only induced a small 14 ± 7% (n = 7) inhibition at +60 mV, which increased to 29 ± 7% (P < 0.05) on hyperpolarization to −40 mV. Thus, Rosc preferentially blocks open HERG channels with a significantly lower affinity for inactivated HERG channels.

Fig. 5.

Rosc blocks open, but not inactivated, HERG channels. A: HERG channels were activated by a 1-s step to +60 mV, and tail current was measured during a 10-s step to −40 mV. Three pulses were given at 1-min intervals between the pulses. During the tail current of the second pulse (*), 30 μM Rosc was rapidly applied and removed for the duration shown by the black bar. The remaining two traces show the current without Rosc application that were recorded 1 min before and 1 min following the Rosc-exposed trace. On average, Rosc blocked the HERG tail current by 38 ± 8% (n = 4). B: current was elicited by an 8-s step to +60 mV, and tail currents were measured at −40 mV. Rosc (30 μM) was rapidly applied during the step and continued into the tail as indicated by the black bar. The superimposed control and recovery currents were recorded 1 min before and after the Rosc-exposed trace, respectively. On average, Rosc block of HERG current at the end of the voltage step to +60 mV was 14 ± 7%, which increased to 29 ± 7% (P < 0.05) on hyperpolarization to −40 mV (n = 7). The arrowheads on the voltage protocol indicate the approximate points at which the step and tail currents were measured. C: the same experiment presented in B was repeated with the inactivation-impaired HERG mutant S620T. The superimposed control current was recorded 1 min before the Rosc-exposed current (*). Unlike the wild-type HERG, S620T currents were equally blocked at +40 mV (step) and −40 mV (peak tail) by application of Rosc (20 ± 6% and 23 ± 6% for step and peak tail currents, respectively, P = 0.45, n = 5). The arrowheads on the voltage protocol indicate the approximate points at which the step and peak tail currents were measured.

If the HERG inactivated state is Rosc insensitive, then equivalent block of step current and tail current should be observed from inactivation-impaired mutants. This hypothesis was tested by measuring the effect of Rosc on the inactivation-impaired HERG mutant S620T (Fig. 5C), which showed large step currents and relatively smaller tail currents (due to lack of inactivation). Rosc was applied in similar manner as shown in Fig. 5B. In contrast to the wild-type HERG, the step current block was not significantly different from that for peak tail current (20 ± 6% and 23 ± 6% for step current and peak tail current, respectively, P = 0.45, n = 5), which supports our hypothesis of similar block for both step and the tail current. Thus, Rosc appears to preferentially block the HERG channel open state with relatively little effect on inactivated HERG channels.

Rapid Rosc-blocking kinetics.

The absence of an effect of Rosc on inactivation kinetics at extreme voltages combined with the smaller block induced by Rosc when applied at +60 mV supports our conclusion that inactivated HERG channels are weakly affected. If this were true, we should be able to modulate the magnitude of block by altering the fraction of inactivated channels. We tested this idea by examining the effect of Rosc on HERG current elicited by a protocol consisting of a 3-s step to +20 mV, followed by a 500-ms repolarization to −40 mV and a subsequent depolarization back to +20 mV for 3 s (Fig. 6A). Under control conditions (in the absence of Rosc), the current increased on repolarization to −40 mV (τ = 13 ± 2 ms, n = 5) and rapidly spiked when the voltage was returned to +20 mV. The amplitude of this spike represents increased driving force (at +20 mV) on K+ moving through the channels that recovered from inactivation at −40 mV, while the rapid decline in current results from inactivation at +20 mV (τ = 8.5 ± 1.2 ms, n = 4). The Rosc effect was calculated from ratio currents generated by (Cntl − Rosc)/Cntl, where Cntl is the average of currents recorded immediately before and on recovery from 30 μM Rosc application (Fig. 6, BD). As predicted, the percent inhibition increased from 26 ± 5% to 37 ± 4% and back to 24 ± 5% for +20, −40 and +20-mV steps (n = 4), respectively (Fig. 6, B and D). We next determined the blocking and unblocking τ from the inhibition time course. The kinetics were not limited by our application system since these measurements were obtained in the constant presence of 30 μM Rosc (applied at least 20 s before initiation of the 3-step voltage protocol). The speed of block (τ = 29 ± 10 ms; n = 4) on stepping from +20 to −40 mV was slower than that of the recovery from inactivation at −40 mV (τ = 13 ± 2 ms, n = 5), which suggests that the channels need to recover from inactivation before Rosc can block. Unblock was also rapid (τ = 22 ± 11, n = 4) when the voltage was stepped back to +20 mV. Given the rapid transition of channels between blocked and unblocked states, there appear to be minimal barriers to Rosc binding and unbinding as HERG channels move between the open and inactivated states.

Fig. 6.

Inactivated channels are insensitive to Rosc. A: HERG currents were activated by a 3-s step to 20-mV, followed by a 1-s hyperpolarization to −40 mV and a return to +20 mV. Representative currents from a single cell are shown. B: the ratio current generated from the currents in A is displayed as percent inhibition by 30 μM Rosc and shows rapid block at the onset of the voltage step from −80 to +20 mV (1); slower unblock during the sustained +20-mV step (2); rapid reblock on hyperpolarization to −40 mV (3); and rapid unblock on returning to +20 mV (4). The ratio current was generated by the formula (Cntl − Rosc)/Cntl, where Cntl is the average of currents before and on recovery from Rosc application. The smooth gray lines are single exponential fits to obtain the τ for initial block, slower unblock during the step to +20 mV, increased block at −40 mV, and decreased block at +20 mV. C: 30 μM Rosc inhibits both step and tail currents (from a representative cell) obtained from inactivation-impaired mutant S620T. D: ratio currents obtained for currents shown in C. Notice that the inhibition increases with current activation and remains relatively stable during the +20-mV step. The arrow marks the repolarization to −40 mV to highlight the absence of the increase of inhibition observed in wild-type HERG.

Rosc does not block closed HERG channels.

In calculating the ratio currents, we noticed that the time course of block at the onset of the first step to +20 mV was surprisingly complex, with the inhibition initially increasing then decreasing (Fig. 6B). Early in the step, block increased from near 0% to a peak of 56 ± 21% with τ = 18 ± 7 ms (n = 9). The block then decreased more slowly to a minimum of 28 ± 8% (τ = 117 ± 44 ms, n = 7) (Fig. 6B). We speculate that the initial increase in inhibition on stepping the voltage to +20 mV results from the rapid block of HERG channels and subsequent decrease in inhibition results from relief of block on inactivation, which is strongly supported by our kinetic model (Fig. 9D). As a direct test of this hypothesis, we ran the same experiment using the inactivation-impaired S620T HERG mutant (Fig. 6, C and D). The ratio currents from the S620T mutant showed the initial rise in the inhibition as observed with wild-type HERG channels, but unlike in the wild type, the inhibition remained relatively stable during the voltage step. Furthermore, there was no change in fractional inhibition on return to −40 mV for currents generated by the S620T mutant (Fig. 6D). The slow decline in the ratio current observed with S620T currents likely results from potassium depletion from the cells during these long voltage steps. This depletion would be stronger for control and recovery currents relative to the inhibited currents in Rosc, which explains the slow decline in apparent inhibition. Large potassium currents are known to deplete small HEK-293 cells of intracellular potassium (19), and we have observed obvious current reductions during 5-s steps to voltages that maximally activate S620T channels. The current reductions were associated with a depolarization of the reversal potential as expected for intracellular potassium depletion. Together, these results demonstrate that Rosc preferentially blocks activated HERG channels with little or no affinity for either closed or inactivated channels.

Rosc is not trapped in the inner vestibule of HERG.

Many HERG blockers like Terf bind in the inner vestibule of the HERG and become trapped by channel closing, which accounts for their slow recovery from block (Fig. 1). To determine whether Rosc can be trapped in closed channels, we hyperpolarized to −120 mV where the deactivation was sufficiently fast, τ = 15 ± 2 ms (80% of channels; n = 4), to capture Rosc if it was binding within the inner vestibule (Fig. 7). (The remaining 20% of channels deactivated with τ = 68 ± 7; n = 4.) However, ratio currents showed the same increase in block with each step in the presence of Rosc, and recovery on Rosc washout was just as rapid as when the deactivation voltage was −40 mV. Thus, Rosc does not appear to be trapped by HERG channel closing.

Fig. 7.

Rosc is not trapped by HERG channel closing at −120 mV. A: peak tail currents were measured at −120 mV following 700-ms voltage steps to +20 mV. The interval between sweeps was 20 s. C (control), R1 (Rosc sweep 1), R2 (Rosc sweep 2), and Re (recovery) indicate sweeps shown in B. B: current traces are shown before, two times during, and on recovery from 30 μM Rosc along with the voltage protocol. C: ratio currents are depicted for both R1 and R2 currents (successive current traces in Rosc). Note the rapid onset of inhibition at the beginning of both R1 and R2, which shows that Rosc unbound during the 20-s interval between sweeps.

HERG channel model reproduces Rosc block.

Our results strongly support preferential block of activated HERG channels by Rosc. A gating model (Fig. 8) based on Wang et al. (42) was developed to further test this mechanism. The rate constant (A, s−1) and charge moved (z) for each transition are given in Table 1. The binding rate constants (k5, k6, and k7) have units of μM−1 s−1. In the diagram shown in Fig. 8, C1–3, O, and I represent closed, open, and inactivated states, respectively; CR, OR, and IR are Rosc-bound closed, open, and inactivated states, respectively. All horizontal transitions are voltage dependent except C2↔C3 (42), and the vertical transitions involve Rosc binding and unbinding.

Fig. 8.

Gating model [based on Wang et al. (42)] of mechanism of preferential block of activated HERG channels by Rosc.

View this table:
Table 1.

Model parameters

We have modified the rate constants from the original model (42) to fit our data and added three Rosc-bound states (CR, OR, and IR; Fig. 8). The major modifications were to increase the charge moved between the open and closed state (C3↔O) to fit the voltage dependence of deactivation τ, and to increase the rate of I→O to fit the recovery from inactivation τ (Table 1; supplemental Fig. 1; the online version of this article contains supplemental data).

We began the model with a single Rosc blocked state that was accessible only from the open state (OR). This was intended to reproduce the apparent open state binding preference of Rosc. However, the main problem with this simple model was that deactivation and inactivation kinetics were drastically slowed by Rosc (not shown), which was not observed experimentally. The slow kinetics resulted from the mass-action movement of Rosc-bound channels into the open state before either inactivating or closing (depending on the voltage), which suggested that additional exits from the Rosc-bound state were needed. The addition of an inactivated blocked state (IR) helped to resolve the problems with inactivation kinetics, while an additional closed blocked state (CB) was required to resolve the problem of roscovitine-induced slowed deactivation. Altering the affinity of the CB state for roscovitine had little affect on the simulated currents so the rate constants were kept the same as O↔OR. However, Rosc binding to the IR state needed to be reduced to match our data (Fig. 5B). This was accomplished by increasing the unbinding rate 5× (IR→I), while the IR→OR rate was increased 5× to maintain microscopic reversibility. The simulated current reproduced our results very well (Fig. 9 and supplemental data Fig. 1). The IC50 for simulated tail current block was 30 μM vs. 27 μM for the experimental data. The block in the model was modulated by voltage as in our experimental results (cf. data in Figs. 3, 4, 6, 9, and supplemental Fig. 1), which arises from both the requirement of channel activation for binding and the decreased affinity of inactivated channels for Rosc. As we observed in our recordings, the inhibition increased with step depolarization during the I-V protocol (Fig. 9B) and also increased with tail hyperpolarization during the tail I-V protocol (supplemental Fig. 1C). The model closely mimicked the kinetics of Rosc block as channels moved between the open and inactivated states (cf. Figs. 6 and 9C). The model ratio currents have a blocking τ = 16 ms (cf. 29 ± 10 ms for our experimental data) for the increased block on a voltage change from +20 to −40 mV, and an unblocking τ = 16 ms (cf. 22 ± 11 ms) on return to +20 mV (Fig. 9D). The model also reproduces the complex block on stepping the voltage from the holding potential to +20 mV (cf. Figs. 6B and 9D). The initial increase in block during activation had a τ = 16 ms (cf. 18 ± 7 ms) and the subsequent decline in block had a τ = 160 ms (cf. 117 ± 44 ms). The model strongly supports our conclusions that Rosc blocks by preferentially binding to activated, but not inactivated, HERG channels. Thus, Rosc therapy may block the open HERG channels (the steady-state window current) that are crucial to proliferation of certain cancers (1, 24).

Fig. 9.

A gating model reproduces Rosc block of HERG channels. A: currents were simulated using a 4-s step to the indicated voltage (x-axis) followed by a 1-s step to −40 mV. Peak tail currents were measured during the −40-mV step and normalized to the current following the +60-mV step. The simulated data are shown for control (▵) and 30 μM Rosc (•). The smooth lines are single Boltzmann equation fits to the experimental data to illustrate the close correspondence with the simulation data. B: the percent inhibition of tail current vs. step voltage by 30 μM Rosc is shown for both simulated (▪) and experimental data (□, reproduced from Fig. 3D). C: the time course of 30 μM Rosc block of simulated current is shown for a 3-s step to +20 mV, which is followed by a 50-ms step to −40 mV and a subsequent return to +20 mV for 3 s. The simulated currents are shown for control (gray trace) and Rosc (black trace). D: the ratio current is shown as percent inhibition and was calculated from the above records in C (cf. Fig 6B).


In this study, we examined the Rosc block of HERG channels using kinetics, mutants, and modeling to determine that Rosc blocks by preferentially binding to activated channels with little or no effect on inactivated channels. Supporting evidence includes increased inhibition with channel activation, rapid block of open channels, and increased block on repolarization where open probability increases following recovery from inactivation (36, 43). The block does not appear to be voltage dependent per se since the inhibition did not change over voltages that generated maximal activation. Instead, the apparent voltage dependence results from the preferential block of activated channels by Rosc. Unlike many other HERG-blocking drugs, inactivated channels are relatively insensitive to Rosc. The block during an inactivating step was significantly less compared with the block during peak tail current (after the channels are recovered from inactivation). However, this difference in block was not observed for currents generated by the inactivation-impaired S620T mutant. The preference of Rosc for activated channels combined with the unique HERG channel inactivation kinetics permitted us to examine the block/unblock process for channels moving between the open and inactivated states (Fig. 6, A and B). This analysis demonstrated the rapid association/dissociation of Rosc with HERG channels, which was reproduced by our model and further supports preferential block of activated HERG channels.

Comparison with other HERG blockers.

Several drugs, including terfenadine, have been removed from the market because they block HERG to induce cardiac arrhythmias (34). These drugs along with class III antiarrhythmic drugs, such as dofetilide, block by binding within the inner vestibule of HERG near the selectivity filter (21, 27, 34). Binding to this site generally results in 1) a preference for these drugs to bind to the open state and 2) a slow recovery from block, both of which results in use-dependent block (17, 37, 40). While Rosc does require channel opening to block, use dependence was not observed, which could result from its rapid block/unblock kinetics. There are some drugs that also rapidly block HERG like spironolactone (τ = 350 ms), fluoxetine (estimated τ = 400 ms), and fluvoxamine (τ = 7.5 ms). However, there are clear differences in the block by these drugs with that induced by Rosc. Fluvoxamine appeared to block closed HERG channels since the majority of channels were blocked at the earliest time point measured (10 ms), and additional data further supported closed channel block (26). Another difference of fluvoxamine block from that induced by Rosc was that inactivated HERG channels also appeared to be blocked. Fluoxetine appeared to block open channels like Rosc, but it had a long (50 min) washout period (38). Spironolactone and its metabolite canrenoic acid blocked rapidly, but like fluvoxamine, appeared to block closed HERG channels (7).

We used the CDK inhibitor Indir to determine whether rapid blocking kinetics were common to all CDK inhibitors. Unlike Rosc, the slower kinetics of the Indir block of HERG current are consistent with kinase inhibition, but further work is required to determine the mechanism. Another kinase inhibitor bisindolylmaleimide I (BIM 1), which inhibits protein kinase C, has been shown to block HERG channels in a kinase-independent manner (39). However, there were several differences between the block induced by BIM I and Rosc. BIM I appeared to block both open and inactivated HERG channels, and recovery from BIM I block was incomplete even after 20 min of washout. In addition, BIM I block was use dependent (39), which we have not observed with Rosc. Therefore, Rosc shows unique preference for open compared with either closed or inactivated HERG channels and rapid blocking kinetics with no use dependence.

Physiological/pathophysiological relevance.

HERG channels are blocked by structurally diverse molecules many of which have been shown to induce acquired long QT syndrome. As a result, many studies have investigated the mechanism of HERG block, and development of molecules that do not block HERG is of considerable interest for drug development (34). However, HERG is a desirable target for block to facilitate treatment of certain cancers (9). We propose that it may be possible to block HERG channels in the cancer cells without significantly increasing the risk for cardiac arrhythmias.

Rosc is currently undergoing phase II clinical trials as an anticancer drug, and the clinically relevant dosing range has been shown to be 10–50 μM (12, 13, 25). However, cardiac arrhythmias have not been reported as a side effect in these patients (3, 18) even though we find the IC50 for HERG block to be ∼30 μM. We believe that rapid Rosc-blocking properties, combined with the inability of HERG channels to trap Rosc, are the reason that cardiac arrhythmias have not been described in clinical studies. These properties allow Rosc to bind and unbind within a single cardiac cycle so that the block is similar for each cAP and does not increase with time (no use dependence). In addition, the absence of use-dependent block means that HERG block will not increase with heart rate, which will help to limit the proarrhythmic potential of Rosc. Furthermore, Rosc has been shown to block cardiac L-channels (6, 44), and it has been reported that L-type calcium channel (CaV1.2) block can compensate for the proarrhythmic effects of HERG block (4, 10, 46). While it is still early in human testing, the absence of arrhythmias is encouraging and suggests that Rosc may block activated HERG channels with minimal cardiovascular side effects.

Potassium channels are of considerable interest as therapeutic targets for cancer chemotherapy (16). HERG channels are overexpressed in many types of neoplastic cell lines and primary tumors (1). The HERG channels in these cells generate a small but steady-state current (window current) that is thought to play a role in proliferation, tumor invasiveness, and tumor neoangiogenesis (1). Blocking the steady-state current could be an effective way to treat these cancers. The combined block of both CDKs and HERG channels (as well as other Kv channels) should enhance Rosc as an anticancer therapy, while the inhibition of L-type calcium currents along with the rapid HERG-blocking kinetics may limit cardiovascular side effects. Therefore, the development of state-specific blockers that can differentially affect HERG current in nonexcitable cancer cells vs. excitable cells, like cardiac myocytes, can offer a unique therapeutic advantage. This could lead to the development of drugs that specifically block HERG channels in transformed (cancer) cells to further minimize potential side effects induced by drug block of HERG channels in nontransformed or excitable cells.


This work was supported in part by an American Heart Association Pre-doctoral fellowship (0715336U) to S. B. Ganapathi; by National Heart, Lung, and Blood Institute Grants RO1-HL-066371 and RO1-HL-076789 to M. Kester; and by a grant from the Pennsylvania (PA) Department of Health using Tobacco Settlement Funds to K. S. Elmslie. The PA Department of Health specifically disclaims responsibility for analyses, interpretations, and conclusions presented here.


The authors thank Drs. Eckhard Ficker (MetroHealth Medical Center, Cleveland, OH) and Shetuan Zhang (Queen's University, Kingston, Ontario, Canada) for the gift of cell lines and clones that were crucial to the completion of this work.


View Abstract