Journal of Neurophysiology

Domain III regulates N-type (CaV2.2) calcium channel closing kinetics

Viktor Yarotskyy, Guofeng Gao, Blaise Z. Peterson, Keith S. Elmslie


CaV2.2 (N-type) and CaV1.2 (L-type) calcium channels gate differently in response to membrane depolarization, which is critical to the unique physiological functions mediated by these channels. We wondered if the source for these differences could be identified. As a first step, we examined the effect of domain exchange between N-type and L-type channels on activation-deactivation kinetics, which were significantly different between these channels. Kinetic analysis of chimeric channels revealed N-channel-like deactivation for all chimeric channels containing N-channel domain III, while activation appeared to be a more distributed function across domains. This led us to hypothesize that domain III was an important regulator of N-channel closing. This idea was further examined with R-roscovitine, which is a trisubstituted purine that slows N-channel deactivation by exclusively binding to activated N-channels. L-channels lack this response to roscovitine, which allowed us to use N-L chimeras to test the role of domain III in roscovitine modulation of N-channel deactivation. In support of our hypothesis, all chimeric channels containing the N-channel domain III responded to roscovitine with slowed deactivation, while those chimeric channels with L-channel domain III did not. Thus a combination of kinetic and pharmacological evidence supports the hypothesis that domain III is an important regulator of N-channel closing. Our results support specialization of gating functions among calcium channel domains.

  • CaV1.2
  • chimera
  • deactivation
  • L-type calcium channel
  • roscovitine

n-type (CaV2.2) calcium channels are complex membrane proteins that consist of one major pore-forming α1B- plus auxiliary β- and α2δ-subunits (Catterall et al. 2005). These channels play a crucial role in central and peripheral nervous systems by controlling neuronal excitability and the release of neurotransmitters and hormones (Harsing et al. 1992; Hirning et al. 1988; Shimosawa et al. 2004; Wheeler et al. 1994). N-channel activity mediates neuropathic pain (Saegusa et al. 2001, 2002), and N-channel-targeted drugs such as ω-conotoxin MVIIA (ziconotide, SNX-111, or Prialt) and ω-conotoxin CVID (AM336) are promising analgesics (Elmslie 2004; McGivern 2006; Snutch 2005; Wermeling and Berger 2006). Despite the crucial function, our understanding of N-channels largely comes from similarity comparison with other voltage-dependent channels.

By analogy with results from work on potassium and sodium channels, it is assumed that the transmembrane (TM) segment S4 moves in response to membrane depolarization to activate voltage-dependent calcium channels (Bezanilla 2000). The voltage sensors appear to be linked to S6, which, in elegant experiments by the Yang group, has been shown to be involved in calcium channel opening and is likely the intracellular activation gate (Xie et al. 2005; Zhen et al. 2005). Thus we have a general idea of how voltage-dependent calcium channels activate and deactivate. However, we do not know how these channels achieve such gating diversity. L-type calcium channels typically gate with brief open times and a low open probability (Po) (maximum ≤ 0.2), while N-type and P/Q-type channels gate with high Po (maximum = 0.8–0.9) and long open times (Colecraft et al. 2001; Elmslie 1997; Forti and Pietrobon 1993; Hess et al. 1984; Lee and Elmslie 1999; Marks and Jones 1992). In addition, L-channels show a voltage-independent open state (Marks and Jones 1992), while the N-channel open state is voltage dependent (Colecraft et al. 2001; Lee and Elmslie 1999). One possible explanation for these gating differences is domain “specialization,” which has been proposed for voltage-dependent sodium channels (Chen et al. 1996; Horn et al. 2000). Sodium channel fast inactivation has been isolated to the intracellular loop between domains III and IV (Patton et al. 1992, 1993), and strong evidence supports domain IV S4 as the specialized voltage sensor for fast inactivation (Chen et al. 1996; Horn et al. 2000). Like sodium channels, we postulate that calcium channels have evolved domain specialization. Indeed, previous studies have demonstrated that activation kinetics of the skeletal muscle L-channel (CaV1.1) can be transferred to the cardiac L-channel (CaV1.2) by CaV1.1 domain I, and vice versa (Tanabe et al. 1991). This suggests that domain I could be an important controller of gating.

We investigated the domain specialization hypothesis by examining activation-deactivation kinetics of chimeric calcium channels comprising domains from N-type and L-type channels, which showed that deactivation kinetics were correlated with domain III. This result was further supported by studies using R-roscovitine, which was recently discovered to slow deactivation of CaV2 (P/Q-, N-, and R-type) channels (Buraei et al. 2007; Yarotskyy and Elmslie 2007; Yarotskyy et al. 2009). R-roscovitine is a trisubstituted purine that was originally developed as a cyclin-dependent kinase (CDK) inhibitor (Meijer 1996). However, the roscovitine-induced slowed deactivation of N-type channels provides previously unavailable opportunities to achieve insights into structures that control channel closing. Since L-channel deactivation is not affected by roscovitine, we used L-N chimeric channels to demonstrate that the N-channel domain III is required for roscovitine-induced slowed deactivation. Our kinetic and pharmacological evidence supports domain III as a primary controller of calcium channel deactivation and suggests that this domain may play a dominant role in generating the high-PO gating characteristic of N-type channels (Colecraft et al. 2001; Lee and Elmslie 1999).


Construction of chimeric channels.

Chimeric calcium channels were constructed with cDNAs encoding the rabbit cardiac L-channel (CaV1.2, GenBank Accession No. X15539) and rat N-channel (CaV2.2, GenBank Accession No. AF055477; generously provided by Dr. Leslie Parent) α1-subunits. Chimeric channels were generated as previously described (Yarotskyy et al. 2010). The overall integrity of each chimera was confirmed by qualitative restriction enzyme digests and DNA sequencing. For convenience, we termed α1-domains by channel type (L or N) and position (I–IV). For example, L-DIII refers to L-type calcium channel domain III. Figure 1 shows a schematic of the chimeric channels used for this study.

Fig. 1.

Chimeric channels construction scheme. Domains I through IV are shown, and each domain is shown as a set of 6 transmembrane segments 1 through 6 (S1–S6). The “+” in each S4 indicates that this segment serves as the voltage sensor. The loops between the segments and domains are shown as lines. N-channel structures are shown in black, while L-channels are in gray. Numbers at the chimeric connections show the number of amino acid residues within that loop contributed by a given channel type. To highlight the connection, the N-channel part of the loop is shown as a thicker line. For the NN*LN chimera, transmembrane S1–S4 in domain II are from the N-channel, while S5–S6 are from the L-channel. The loop between S4 and S5 is from the L-channel.

HEK cell transfection.

We utilized the calcium phosphate precipitation method to exogenously express channels in HEK293 cells (Wang et al. 2005; Yarotskyy et al. 2010). HEK293 cells were maintained in standard DMEM-GlutaMAX medium containing 10% fetal bovine serum (FBS) and 1% antibiotic-antimycotic (regular medium) at 37°C in a 5% CO2 incubator. Cells were transfected with 11 μg of α1-subunit (L-, N-, or chimeric channel), 8.5 μg of α2δ, 5.5 μg of β1b, 2.15 μg of TAG (SV40 large T-antigen, to increase expression efficiency), and 1 μg of green fluorescent protein (GFP; to visualize transfected cells). The transfected cells were split into 35-mm dishes that served as the recording chamber.

Measurement of ionic currents.

Cells were voltage clamped with the whole cell configuration of the patch-clamp technique. Pipettes were pulled from Schott 8250 glass (King Precision Glass, Claremont, CA) on a Sutter P-97 puller (Sutter Instruments, Novato, CA). Currents were recorded with an Axopatch 200A amplifier (Molecular Devices, Sunnyvale, CA) and digitized with ITC-18 data acquisition interface (Instrutech, Port Washington, NY). Experiments were controlled by a Power Macintosh G3 computer (Apple Computer, Cupertino, CA) running S5 data acquisition software written by Dr. Stephen Ikeda (National Institute on Alcohol Abuse and Alcoholism, National Institutes of Health, Bethesda, MD). Leak current was subtracted online with a −P/4 protocol. The impact of series resistance (corrective) and whole cell capacitance (predictive) on the whole cell voltage clamp was compensated by at least 80% with circuitry of the Axopatch 200A. All recordings were carried out at room temperature, and the holding potential was −120 mV. Whole cell currents were digitized at 50 kHz after analog filtering at 5–10 kHz.

Data analysis.

Data were analyzed with IgorPro (WaveMetrics, Lake Oswego, OR) running on a Macintosh computer. The voltage dependence of channel activation was measured from tail currents by averaging for 0.3 ms beginning 0.3 ms after repolarization to −60 mV from voltages ranging from −80 to 80 mV. The tail current amplitude was plotted versus step voltage and fitted by a Boltzmann function, which yielded half-activation voltage (V1/2) and slope factor. Activation and deactivation kinetics were determined with single-exponential fitting as previously described (Buraei et al. 2005). Group data are calculated as means ± SD throughout this report. Paired t-test was used for within-cell comparisons. One-way ANOVA with Tukey honestly significant difference (HSD) post hoc test was used to test for differences among three or more independent groups. Significant differences required P < 0.05.


The internal pipette solution contained (in mM) 104 NMG-Cl, 14 creatine-PO4, 6 MgCl2, 10 NMG-HEPES, 5 Tris-ATP, 0.3 Tris-GTP, and 10 NMG-EGTA, with osmolarity of 280 mosM and pH of 7.3. The external recording solution contained (in mM) 30 BaCl2, 100 NMG-Cl, and 10 NMG-HEPES, with osmolarity of 300 mosM and pH of 7.3. The mean liquid-liquid junction potential was −2.7 mV, and we did not compensate for this potential in our analysis. Importantly, the internal and external solutions were identical for all recordings so that the junction potential was the same for all chimeric channels examined. The 30 mM Ba2+ concentration in the external solution was used to ensure that current could be measured from all chimeric channels under the same conditions, since some chimeras generate small currents. Higher divalent cation concentrations induce a depolarizing shift in voltage-dependent properties (Elmslie et al. 1994; Liang and Elmslie 2001), but channel activation and deactivation kinetics are normal after adjusting for this voltage shift (Zhou and Jones 1995). R-roscovitine was prepared as a 50 mM stock solution in DMSO and stored at −30°C. All external solutions contained the same DMSO concentration so that the roscovitine concentration was the sole variable when changing solutions. Test solutions were applied from a gravity-fed perfusion system with an exchange time of 1–2 s.


R-roscovitine was from LC Labs (Woburn, MA). DMEM, FBS, and antibiotic/antimycotic were from Invitrogen (Carlsbad, CA). Other chemicals were obtained from Sigma (St. Louis, MO).


N-channel-like deactivation follows domain III.

It has long been known that N-channels show several unique gating characteristics relative to L-channels such as high Po and long open times (Lee and Elmslie 1999; Marks and Jones 1992). In addition, N-channels show voltage-dependent open times, while L-channel open times do not change with voltage (Lee and Elmslie 1999; Marks and Jones 1992). Since the mechanisms that underlie these gating differences are not understood, we wondered if N-L chimeras could be used to gain insights into these differences. As a first step toward understanding the root of these differences, we examined whole cell current activation and deactivation kinetics to determine whether certain kinetic properties could be isolated to single domains. For wild-type (wt) N- and L-channels, both activation and deactivation time constants (τ) were different at certain voltages, but differences at single voltages are highly susceptible to many factors that make them unreliable for our purposes. For this reason, we compared these values over a range of voltages to determine the e-fold change with voltage (Ve) (Buraei et al. 2005). The Ve for both activation and deactivation τ was significantly smaller (i.e., steeper voltage dependence) for wt N-type vs. L-type channels (Fig. 2). For activation τ averaged from six cells for N-channels and seven cells for L-channels, VeAct Ve) was 34 mV for N-channels versus 125 mV for L-channels, whereas deactivation τ VeDeact Ve) from fitting averaged results (n = 5 for N-channels and n = 5 for L-channels) yielded 28 mV for N-channels versus 47 mV for L-channels (Fig. 2). Statistical analysis of τAct Ve and τDeact Ve (averaged from fitting single cells) showed that both values significantly differed between N-type and L-type channels (see Fig. 4). The steeper voltage dependence of N-channel τDeact likely reflects the voltage dependence of the N-channel open state (Colecraft et al. 2001; Lee and Elmslie 1999).

Fig. 2.

Comparison of the whole cell kinetics of N- and L-current. N- and L-channels were transiently expressed in HEK293 cells along with β1b- and α2δ-subunits. A: example currents were recorded in 30 mM Ba2+ from cells expressing either L (LLLL)- or N (NNNN)-channels. Currents were activated during 25-ms steps to +20 mV, and tail currents were measured at −60 mV. B: the time constant of activation (τAct) was determined by fitting activation with a single-exponential equation after a 0.3-ms delay and is plotted vs. step voltage. The smooth lines are single-exponential fits to determine the e-fold change in τ with voltage (Ve). NNNN τAct (n = 6) changed e-fold for 34 mV (black line), and LLLL τAct (n = 7) changed e-fold for 125 mV (gray line). C: the time constant of deactivation (τDeact) was determined by fitting tail currents to a single-exponential equation and is plotted vs. tail voltage. The smooth lines were generated as described in B. NNNN τDeact (n = 5) changed e-fold for 28 mV (black line), and that for LLLL (n = 5) changed e-fold for 47 mV (gray line).

LLNN was the first chimera we tested to determine whether swapping half of each channel would produce gating changes that would align with either wt N- or L-channels. The LLNN chimera showed activation and deactivation kinetics that were similar to those of the wt N-channel and were significantly different from those of wt L-channels (see Figs. 1, 3, and 4), which focused our attention on N-DIII and N-DIV. As we have noted previously, chimeric channels that carried N-DII (including NNLL) failed to generate current (Yarotskyy et al. 2010). Fortunately, all chimeric channels with L-DII were functional, which allowed us to test the role of N-DIII and/or N-DIV in activation-deactivation. L-channel chimeras containing one or two N-channel domains showed that τDeact Ve was statistically different between chimeras containing N-DIII (NNNN, LLNN, and LLNL) and L-DIII (LLLL, NLLL, and LLLN) (Fig. 3, 4B). Thus the inclusion of N-DIII into L-channel backbone (LLNL) made τDeact Ve N-channel-like, while the LLLN chimera showed L-channel-like τDeact Ve. Interestingly, τAct Ve was significantly different from that of the wt L-channel for all chimeric channels tested (Fig. 4A). Activation kinetics are more influenced by closed-closed transition rates compared with deactivation kinetics (Yarotskyy and Elmslie 2009), which suggests that the insertion of any N-channel domain can increase transition rates among closed states. However, another measure of activation was strongly affected by N-DIII and N-DIV. N-channels (NNNN) show a significantly steeper activation voltage dependence (Boltzmann slope factor) that was only reproduced in the LLNN chimera (Fig. 4D). The Boltzmann half-activation voltage (V1/2) was not correlated with any single domain switch between N-type and L-type channels, which likely results from this parameter being sensitive to both closed state vs. open state stability and the voltage dependence of activation/deactivation (Logothetis et al. 1993).

Fig. 3.

Typical currents recorded from chimeric channels. All panels comprise the current example (top) and a voltage record (bottom). Vertical and horizontal lines show scales for current and time, respectively. Currents are shown for LLNN (A), LLNL (B), NN*LN (C), LLLN (D), and NLLL (E) channels.

Fig. 4.

The N-channel domain III dictates deactivation gating. A: τAct Ve was determined as described in Fig. 2. All channels containing N-channel domains showed a significantly different τAct Ve compared with that from wild-type (wt) L-type channels (LLLL). B: τDeact Ve values for NNNN, LLNN, and LLNL were significantly different from LLLN, NLLL, NN*LN, and LLLL channels. Half-activation voltage (V1/2, C) and slope (D) were determined from single Boltzmann equation fits to the activation current-voltage relationship (I-V) as described in Fig. 5. V1/2 was similar for all channels except the LLNL and LLLN chimeras. Significant differences for all comparisons were determined by ANOVA with Tukey honestly significant difference (HSD) post hoc test (P < 0.05). Lowercase letters above or below each column indicate the data that differ significantly with NNNN (a), LLNN (b), LLNL (c), LLLN (d), NLLL (e), NN*LN (f), and LLLL (g). The number of cells tested is shown for each column.

These results supported N-DIII as the mediator of the steeper voltage dependence of N-channel closing. We wanted to further test this idea by measuring the kinetic parameters of the NNLN chimera, which we expected to show L-type τDeact Ve. However, the NNLN chimera contained N-DII, which prevented the expression of functional channels (Yarotskyy et al. 2010). In an effort to isolate the N-DII structures involved in channel dysfunction, we generated hemidomain channels containing domain II with N-type S1–S4 and L-type S5–S6 (Fig. 1), which we found could generate functional channels (Yarotskyy et al. 2010). This hemidomain construct is designated N* throughout, and it allowed us to make NN*LN to test our prediction. Indeed, we found that the τDeact Ve of this chimera was statistically similar to that of wt L-channels, which supports the idea that domain III may be specialized to control calcium channel closing. τAct Ve for this chimera was statistically similar to N-channels (Fig. 4A), which further supports the idea that the control of activation may be more distributed among the domains. The Boltzmann slope factor for NN*LN was similar to that of the L-type channels, which further supports a role for both N-DIII and N-DIV in establishing the steeper voltage dependence of activation observed for wt N-channels (Fig. 4D). Together our results support the idea that domain III may be specialized to control calcium channel closing and, thus, contribute to the gating differences between L-type and N-type channels.

The α2δ-subunit fails to modulate N-channel closing.

Under control conditions, N-channels rapidly deactivate upon hyperpolarization. However, roscovitine, a trisubstituted purine, can slow deactivation of CaV2 channels by stabilizing high-Po open state (Buraei et al. 2005, 2007). This effect was accompanied by an increase in the τDeact Ve, left shift in the activation vs. voltage relationship (ΔV1/2) (Buraei et al. 2005, 2007), and slowed gating charge relaxation (Yarotskyy and Elmslie 2009). Thus roscovitine provides us with another method to probe structures that control N-channel closing. The requirement of CaV2 channel activation for the roscovitine-induced slowed deactivation, as well as the fact that externally but not internally applied roscovitine induced the effect, suggested an externally oriented binding site (Buraei et al. 2005, 2007). While we had previously proposed that the binding site resided on the α1B-subunit, the associated α2δ-subunit is extracellularly exposed, and is thought to be the target for clinically relevant compounds such as gabapentin to affect channel activity (Brown and Gee 1998). Thus, before testing roscovitine on chimeric calcium channels, we needed to show that the relevant site was not on the α2δ-subunit, which was tested by expressing α1B- and β1b-subunits in HEK293 cells with and without α2δ. The absence of the α2δ-subunit failed to affect any of the monitored N-channel gating parameters. The activation vs. voltage relationship V1/2 was 19.8 ± 7.1 mV (n = 14) vs. 22.7 ± 4.1 [n = 4, not significantly different (ns)] and the Boltzmann slope factor was 10.9 ± 2.3 vs. 12.3 ± 3.1 (ns) with or without α2δ, respectively. The τAct Ve was 31.2 ± 12.3 (n = 12) vs. 38.8 ± 14.6 (n = 5, ns) and τDeact Ve was 28.7 ± 3.3 (n = 11) vs. 31.0 ± 6.3 (n = 5, ns) with and without α2δ, respectively. The absence of the α2δ-subunit also had no effect on roscovitine-induced slowed deactivation (Fig. 5). Roscovitine at 100 μM induced a shift in V½, with the average ΔV1/2 = −8.2 ± 5.5 mV with α2δ (n = 14) and −6.5 ± 3.2 without α2δ (n = 4, ns), and τDeact Ve was significantly increased with roscovitine relative to control, with the percent change in τDeact Ve = 152 ± 56% with α2δ (n = 11) and 196 ± 56% without α2δ (n = 5, ns). These results show that the α2δ-subunit does not impact the roscovitine-induced slowed deactivation and support the idea that roscovitine binds to the N-channel α1-subunit to affect deactivation. Thus testing roscovitine on the chimeric channels will provide further insights into the domain(s) that control N-channel closing.

Fig. 5.

Roscovitine-induced slowed deactivation results from interaction with the α1B-subunit. Data are shown from HEK293 cells expressing α1B-, α2δ-, and β1b-subunits (+α2δ, left) and α1B- and β1b-subunits (−α2δ, right). A: 100 μM roscovitine induced a left shift in the activation I–V for both groups. The data were fitted by a single Boltzmann function (smooth line) to determine V1/2 and Boltzmann slope factor. Tail current (ITail) amplitudes were scaled (normalized) by the maximum value from the Boltzmann equation fit. Boltzmann fitting values for control (Cntl), 100 μM roscovitine (Rosc), and washout (WO) were as follows: +α2δ: V1/2 = 19.7, 4.7, 20.6 mV and slope = 12.8, 8.0, 12.5; −α2δ: V1/2 = 27.6, 8.7, 29.5 mV and slope = 11.5, 8.7, 11.2, respectively. B: roscovitine slowed deactivation of tail currents and increased the τDeact Ve. Tail currents were fit by a single-exponential function to obtain τDeact for each tail voltage. The plot of τDeact vs. voltage was fit by using a single-exponential equation to determine the e-fold change in τDeact with voltage (Ve). τDeact Ve for Cntl, Rosc, and WO were as follows: +α2δ: −25.9, −77.4, −26.2 mV; −α2δ: −23.8, −80.9, −24.6 mV, respectively. C: representative traces show the roscovitine-induced slowed deactivation (black traces) compared with control and washout (gray traces) for HEK293 cells expressing αlB-subunits with or without α2δ-subunits.

N-DIII mediates roscovitine-induced slowed deactivation.

We have previously established that roscovitine failed to slow L-channel deactivation (Buraei et al. 2007), which allowed us to use our chimera strategy to determine the N-channel domain(s) that mediate slowed deactivation. To facilitate comparisons with the chimeric channels, the effect of roscovitine on L-channel gating is shown in Fig. 6A, and Fig. 5 shows the effect on wt N-channel gating. The LLNN chimera was used to localize the agonist effect to half of the channel. The current generated by this chimera clearly shows roscovitine-induced slowed deactivation along with an increased τDeact Ve, which completely contrasts with the effect of roscovitine on the wt L-channel (Fig. 6). The roscovitine-induced change in τDeact Ve of the LLNN chimera was significantly different from that of wt L-channels but statistically similar to that of wt N-channels (see Fig. 8A). Interestingly, roscovitine failed to shift the activation-voltage relationship in the LLNN chimera (Fig. 6B; see Fig. 8B). As discussed below, this likely results from the additional effects of roscovitine on the L-channel half of the chimera (Yarotskyy et al. 2010). These results localize the structures involved in roscovitine-induced slowed deactivation to N-DIII and/or N-DIV.

Fig. 6.

Roscovitine-induced slowed deactivation is mediated by N-DIII. Left: L-channel data (LLLL). Center and right: data from LLNN and LLLN chimeras, respectively. Please see Fig. 5 for comparative wt N-channel data. A: the activation I-V was not shifted by 100 μM Rosc for either LLLL or LLNN channels, while for LLLN there was a small positive shift. V1/2 and Boltzmann slope factor were obtained as described in Fig. 5. Values for CNTL, 100 μM Rosc, and WO were as follows: LLLL: V1/2 = 7.9, 6.6, 9.1 mV and slope = 12.1, 9.9, 12.9; LLNN: V1/2 = 9.1, 8.2, 8.3 mV and slope = 9.9, 8.5, 8.0; LLLN: V1/2 = 45.0, 48.3, 43.3 mV and slope = 16.7, 16.1, 15.0, respectively. B: roscovitine slowed deactivation of the LLNN tail current but not that of either wt L (LLLL) or LLLN channels. Measurements were done as described in Fig. 5. τDeact Ve for CNTL, Rosc and WO were as follows: LLLL: 31.1, 31.7, and 35.0 mV; LLNN: 17.9, 56.8, and 18.1 mV; LLLN: 35.5, 42.0, and 39.8 mV, respectively. Symbols have the same meaning as in A. C: current traces show the slowed activation of LLLL channels (left) and slowed deactivation of the currents from the LLNN chimera (center), but neither of these effects was observed for LLLN chimera. Current traces in the presence of 100 μM Rosc are shown in black, while those of CNTL and WO are in gray.

Single domain substitutions were used to further localize the structures involved with slowed deactivation. Figures 6 and 7 clearly illustrate that the LLNL chimera shows an N-channel-like response to roscovitine, while the LLLN chimera does not. Deactivation of LLNL channels was slowed by application of 100 μM roscovitine, which resulted in a significant increase in τDeact Ve from 24.8 ± 1.9 mV (average of control and washout) to 55.3 ± 5.2 mV (P < 0.01, n = 5; Fig. 7A). The roscovitine-induced change in τDeact Ve for the LLNL chimera was significantly larger than that for wt L-channels and chimeric channels that lacked N-III (Fig. 8A). However, the percent change in τDeact Ve was significantly smaller than that for either wt N-channels or the LLNN chimera (Fig. 8A). In contrast, roscovitine failed to affect deactivation of the LLLN chimera (Fig. 6, Fig. 8A) and τDeact Ve was not significantly different between control (43.3 ± 9.0 mV; average of control and washout) and roscovitine (45.4 ± 7.7 mV; n = 6) for this chimera. Surprisingly, roscovitine significantly right-shifted the activation vs. voltage relationship of the LLNL chimera (ΔV1/2 was 14.8 ± 1.8 mV, P < 0.001, n = 4) (Fig. 7A, Fig. 8B), which was unique among all the channels tested. This effect likely results from the effect of roscovitine on the L-channel domains within the chimera, namely, slowed activation (Yarotskyy et al. 2010).

Fig. 7.

The N-channel domain III is necessary and sufficient for roscovitine-induced slowed deactivation. LLNL chimera data are presented on left, whereas those from the NN*LN chimera are shown on right. A: the activation I–V was right-shifted by application of 100 μM roscovitine for the LLNL chimera. V1/2 and Boltzmann slope factor were determined as described in Fig. 5 for control (Cntl), 100 μM roscovitine (Rosc), and washout (WO): LLNL: V1/2 = −17.6, −2.6, −21.1 mV and slope = 12.1, 13, 10.5; NN*LN: V1/2 = 6.2, 4.5, 5.4 mV and slope = 16.5, 13.1, 16.2, respectively. B: roscovitine slowed deactivation of the LLNL chimera but not that of NN*LN channels. τDeact Ve in control, roscovitine, and washout were 24.9, 52.3, 25.8 mV for the LLNL chimera and 35.4, 36.9, and 31.8 for the NN*LN chimera, respectively. C: current traces in the presence of 100 μM roscovitine are shown in black, while those of control and washout are in gray. Note that roscovitine induces both slowed activation and slowed deactivation of currents from the LLNL chimera.

Fig. 8.

Roscovitine-induced increase in τDeact Ve depends on N-channel domain III. A: the 100 μM roscovitine-induced % change in τDeact Ve was determined by averaging the Ve values from control and washout before calculating the % change. B: ΔV1/2 was determined by averaging V1/2 values from control and washout before subtracting that measured in 100 μM roscovitine. The significant differences for all comparisons were determined by ANOVA with Tukey HSD post hoc test (P < 0.05). Lowercase letters above or below each column indicate the data that differ significantly from NNNN (a), LLNN (b), LLNL (c), LLLN (d), NLLL (e), NN*LN (f), and LLLL (g). The number of cells tested is shown for each column.

We also examined the NLLL chimera and found that roscovitine failed to affect deactivation as expected from our results from the LLNN chimera (Fig. 8A). To support the idea that N-DIII is both necessary and sufficient for roscovitine-induced slowed deactivation, we tested the NN*LN chimera (Fig. 7B). As we predicted, 100 μM roscovitine failed to affect deactivation of the NN*LN channel (Figs. 7 and 8). Thus we conclude that domain III controls N-channel closing.


As a first step toward understanding the structures that generate the kinetics differences between N-type and L-type channels, we utilized N-L chimeras to determine whether certain kinetic parameters could be isolated to single domains. Analysis of the kinetic properties of these channels demonstrated that the N-channel characteristic of steeper τDeact voltage dependence was associated with N-DIII, whereas the steeper voltage dependence of activation (Boltzmann slope factor) appeared to require both N-DIII and N-DIV. The difference in τAct Ve between N-type and L-type channels was not correlated with any domains and thus appears to be a more distributed function. We and others have previously demonstrated that roscovitine slows deactivation of P/Q-, N-, and R-type calcium channels, which results in elevated action potential-induced calcium influx (Buraei et al. 2005, 2007; Cho and Meriney 2006; Yan et al. 2002). We show here that N-DIII is both necessary and sufficient to mediate this effect. Together these results suggest that the calcium channel DIII is an important controller of deactivation and imparts at least some of the unique kinetic characteristics observed between N-type and L-type channels.

Domain specialization.

Most of our knowledge of the structures involved in ion channel gating comes from potassium channels, which comprise four independent α-subunits to form the channel (MacKinnon 1991). Thus each α-subunit must carry all gating functions (i.e., voltage sensing, activation gating, and inactivation gating), and each subunit appears to contribute to channel activation (Horn et al. 2000) and inactivation (Hoshi et al. 1990; Zagotta et al. 1990). On the other hand, sodium and calcium channels have all four domains together in one protein (Catterall 2000), which could allow for specialization. Over the past decade evidence has accumulated supporting domain specialization in voltage-dependent sodium channels. Sodium channel fast inactivation has been isolated to the intracellular loop between domains III and IV (Patton et al. 1992, 1993), and the sodium channel DIV-S4 appears to be the voltage sensor for fast inactivation (Chanda and Bezanilla 2002; Chen et al. 1996; Horn et al. 2000; Yang and Kuo 2003). Thus the domains of sodium and calcium channels may have evolved to contribute unique functions, and the identification of a voltage-dependent calcium channel in yeast (CCH1) suggests that there has been sufficient time to undergo such evolution (Paidhungat and Garrett 1997).

There are several differences between N-type and L-type channel kinetics that we could discern from whole cell currents. We wanted robust measures of channel function and so rejected measurements taken at a single voltage in favor of measurements taken across a range of voltages, which we believed would better reflect the underlying gating processes. One such measurement was τAct Ve, which was significantly smaller (i.e., exhibited steeper voltage dependence) for N-type vs. L-type channels. However, we were unable to attribute this difference to a single domain or even two domains, since any domain change induced a significant decrease in τAct Ve compared with that of wt L-channels. This is not surprising since this measure is very sensitive to closed state transitions (Buraei et al. 2005; Marks and Jones 1992). Voltage-dependent channels move through multiple closed states on the pathway to opening, and each of these voltage-dependent steps can have an impact on τAct. It is surprising that replacing any one of the L-channel domains resulted in a significant change in τAct Ve. It seems that all L-channel domains are involved in establishing the weaker voltage dependence of τAct.

It was previously demonstrated that differences in activation τ were determined by DI of the skeletal muscle (CaV1.1) and cardiac (CaV1.2) L-channels (Tanabe et al. 1991). Activation of CaV1.1 is an order of magnitude slower than that of CaV1.2. Chimeric channels containing CaV1.2 DI activated with a τ that was statistically similar to that of wt CaV1.2, while those channels with CaV1.1 DI activated significantly slower. Thus DI of the L-channel appears to control activation kinetics (Tanabe et al. 1991). However, the voltage dependence of τAct was not examined in that study. One possibility is that DI of CaV1.1 introduces a rate-limiting step to activation, causing channels with this domain to activate slowly. In our case, activation kinetics were not different between N-type and L-type channels over a large range of voltages, so we focused on the voltage dependence of activation in our studies.

Another significant difference between N-type and L-type channels was the steeper steady-state activation vs. voltage-relationship for N-channels. One might expect that this parameter would be very sensitive to changes in both closed-closed and closed-open gating, which would suggest that a single domain would have little impact. While our data showed that introduction of a single N-channel domain into the L-channel backbone failed to significantly alter the Boltzmann slope factor, the introduction of L-DIII into the N-channel (NN*LN) had a significant effect to decrease the steepness of activation to that of the L-channel, which suggests that N-DIII is important for this critical channel gating parameter. However, N-DIII appears to require N-DIV to mediate steeper activation, since only the LLNN chimera showed a slope factor that was significantly different from that of the wt L-channels and statistically similar to that of the wt N-channel. Thus these two domains may together contribute to increased sensitivity of N-channel activation to membrane depolarization. Interestingly, each of these two N-channel domains carries one additional positively charged amino acid compared with the corresponding L-channel domains, which could explain the steeper voltage dependence of N-channel activation.

Our chimera data show that DIII is crucial to establish the voltage dependence of calcium channel deactivation, since the N-like τDeact Ve required N-DIII. The steeper voltage dependence of N-channel deactivation very likely reflects the voltage-dependent N-channel open state. Our modeling shows that open state transitions have a strong control over N-channel deactivation kinetics (Buraei et al. 2005; Yarotskyy and Elmslie 2009). This suggests that DIII may be a primary controller of calcium channel Po, which is highly dependent on the transition rate from the open to the closed state (Wonderlin et al. 1990).

One potential mediator of this voltage dependence to the N-channel open state is charged amino acids within N-DIII-S4, which has the highest number of positively charged amino acids (6) relative to any L-channel domain (4–5) or other N-channel domains (5). Perhaps the extra charge mediates the increased τDeact Ve observed in channels with N-DIII. Further investigation of domain III could uncover the structural mechanisms that distinguish the gating of N-type from L-type calcium channels.

N-DIII is critical for roscovitine-induced slowed deactivation.

Roscovitine was initially developed as a CDK inhibitor (Meijer 1996) that has more recently been used to provide unique insights into N-channel gating (Buraei et al. 2005; Buraei and Elmslie 2008; DeStefino et al. 2010; Yarotskyy and Elmslie 2009, 2010). The profound slowing of N-channel closing (Buraei et al. 2005) induced by this drug provided a good test for our hypothesis that DIII controls N-channel closing.

Experiments on the N-L chimeric channels showed that roscovitine slowed deactivation in all channels containing N-DIII (NNNN, LLNN, and LLNL). The LLNL and NN*LN chimeras are particularly important since LLNL shows that N-DIII is sufficient to transfer the roscovitine-induced slowed deactivation to L-type channels, while NN*LN demonstrates that N-DIII is necessary for slowed deactivation of N-type channels. The simplest explanation that accounts for our kinetic and roscovitine results is that DIII contains both the roscovitine binding site and the gating machinery required for channel closing. However, it is also possible that the roscovitine binding site is located elsewhere on the channel and is conserved between L- and N-channels. One interesting difference we observed is that the roscovitine-induced %ΔτDeact Ve was significantly smaller for LLNL relative to either LLNN or the wt N-channel, suggesting that N-DIV is required for a full roscovitine response. However, it is clear from our data that N-DIV does not mediate roscovitine responsiveness. This observation suggests that the gating machinery in N-DIII is energetically coupled to molecular determinants in N-DIV, but N-DIII alone is necessary and sufficient for roscovitine-dependent changes in deactivation. Together our kinetic analysis of the N-L chimeric channels and of the roscovitine effect on those channels supports the hypothesis that DIII is a critical controller of N-channel deactivation, and reveals for the first time domain specialization for the gating of N-type calcium channels.

We have previously demonstrated that roscovitine can induce N-channel inhibition as well as slowed deactivation (Buraei et al. 2005, 2007). This N-channel inhibition results from enhancement of closed-state inactivation (Buraei and Elmslie 2008), and the inhibitory effect appears to require N-DI (Yarotskyy et al. 2010). Thus it is likely that two roscovitine binding sites exist on the N-channel. One site in N-DIII regulates N-channel closing, and another in N-DI regulates closed-state inactivation.

The BayK8644-induced slowing of L-channel deactivation is superficially similar to the roscovitine effect on N-channels. BayK8644 binds to a site comprising both L-DIII and L-DIV, and the site appears to be accessible via the membrane (Hockerman et al. 1997). The available evidence supports the roscovitine-binding site within N-DIII, and we have argued that the site is extracellularly exposed (Buraei et al. 2005, 2007). Thus it seems unlikely that the roscovitine site is evolutionarily related to the BayK8644 site. In addition, there are significant differences in the effects of BayK8644 and roscovitine that further support this idea. First, roscovitine exclusively binds activated N-channels to slow deactivation (Buraei et al. 2005), while BayK8633 appears to bind to closed channels to affect gating (Hockerman et al. 1997). Second, roscovitine slows off-gating charge movement of N-type channels (Yarotskyy and Elmslie 2009), while drugs that slow L-channel deactivation such as BayK8644 and FPL64176 have little or no effect on gating charge movement (Artigas et al. 2003; Fan et al. 2000; McDonough et al. 2005). Thus the evidence supports a strong distinction between the effects of roscovitine on N-channels and BayK8644 on L-channels.

Clinical relevance of slowed deactivation.

CaV2 channels are localized within nerve terminals in both the central and peripheral nervous systems and serve as the major pathway for Ca2+ entry that triggers neurotransmitter release (Harsing et al. 1992; Hirning et al. 1988; Shimosawa et al. 2004; Wheeler et al. 1994). Until the identification of R-roscovitine, there were no drugs available to enhance Ca2+ influx through presynaptic CaV2 channels (Buraei et al. 2005, 2007; Yan et al. 2002). Several studies have demonstrated that R-roscovitine can enhance neurotransmitter release (Cho and Meriney 2006; Yan et al. 2002) and, thus, could be used as a treatment for diseases that result from either reduced neurotransmitter release (e.g., Lambert-Eaton syndrome and Parkinson's disease) or diminished neurotransmitter receptor levels (e.g., myasthenia gravis). One potential problem with using R-roscovitine is cross-reactivity with other proteins including kinases, voltage-dependent potassium channels, and L-type calcium channels. Thus identification of the binding site could provide the information needed for the structure-directed synthesis of roscovitine analogs that have enhanced specificity for the agonist-binding site on CaV2 channels. The progress toward identification of this site has highlighted a previous unknown functionality of DIII as a regulator of N-channel deactivation.


This work was supported in part by grants from the Pennsylvania (PA) Department of Health using Tobacco Settlement Funds and National Institutes of Health Grants AR-059397 (K. S. Elmslie) and HL-074143 (B. Z. Peterson). The PA Department of Health specifically disclaims responsibility for analyses, interpretations, and conclusions presented here.


No conflicts of interest, financial or otherwise, are declared by the author(s).


Author contributions: V.Y., G.G., B.Z.P., and K.S.E. conception and design of research; V.Y., G.G., and K.S.E. performed experiments; V.Y., G.G., and K.S.E. analyzed data; V.Y., G.G., B.Z.P., and K.S.E. interpreted results of experiments; V.Y. and K.S.E. prepared figures; V.Y., B.Z.P., and K.S.E. drafted manuscript; V.Y., G.G., B.Z.P., and K.S.E. edited and revised manuscript; V.Y., G.G., B.Z.P., and K.S.E. approved final version of manuscript.


We thank Lei Du and Yunhua Wang for superb technical assistance in chimeric channel development and preparation.

Present addresses: V. Yarotskyy, Dept. of Pharmacology and Physiology, University of Rochester, 601 Elmwood Ave., Rochester, NY 14642; Guofeng Gao, Dept. of Medicine, Penn State College of Medicine, Penn State University, Hershey, PA 17033.


View Abstract