New Cav2 calcium channel gating modifiers with agonist activity and therapeutic potential to treat neuromuscular disease

Abstract

Voltage-gated calcium channels (VGCCs) are critical regulators of many cellular functions, including the activity-dependent release of chemical neurotransmitter from nerve terminals. At nerve terminals, the Cav2 family of VGCCs are closely positioned with neurotransmitter-containing synaptic vesicles. The relationship between calcium ions and transmitter release is such that even subtle changes in calcium flux through VGCCs have a strong influence on the magnitude of transmitter released. Therefore, modulators of the calcium influx at nerve terminals have the potential to strongly affect transmitter release at synapses. We have previously developed novel Cav2-selective VGCC gating modifiers (notably GV-58) that slow the deactivation of VGCC current, increasing total calcium ion flux. Here, we describe ten new gating modifiers based on the GV-58 structure that extend our understanding of the structure-activity relationship for this class of molecules and extend the range of modulation of channel activities. In particular, we show that one of these new compounds (MF-06) was more efficacious than GV-58, another (KK-75) acts more quickly on VGCCs than GV-58, and a third (KK-20) has a mix of increased speed and efficacy. A subset of these new VGCC agonist gating modifiers can increase transmitter release during action potentials at neuromuscular synapses, and as such, show potential as therapeutics for diseases with a presynaptic deficit that results in neuromuscular weakness. Further, several of these new compounds can be useful tool compounds for the study of VGCC gating and function.

Graphical abstract

1. INTRODUCTION

Chemical communication in the nervous system is controlled by nerve terminal voltage-gated calcium channels (VGCCs), particularly the Cav2 family of VGCCs that are selectively expressed at sites of chemical communication. The calcium ion flux that enters through these channels into the nerve terminal provides the biochemical trigger for synaptic vesicle fusion and the release of chemical transmitters (Katz and Miledi, 1965; Katz, 1969). Further, this calcium flux is non-linearly related to the magnitude of chemical transmitter release such that small changes in calcium flux produce very large changes in transmitter release (Dodge and Rahamimoff, 1967). As such, changes in the number of VGCCs at synapses, or in VGCC gating are critical determinants of the strength of communication between cells in the nervous system. This is particularly relevant in the neurological disease Lambert-Eaton myasthenic syndrome (LEMS) which is characterized by an autoimmune-mediated reduction in the number of VGCCs at motor nerve terminals (Lambert et al., 1956; Lennon et al., 1995; Motomura et al., 1997; Nagel et al., 1988; Vincent et al., 1989; Meriney et al., 1996). This autoimmune-mediated reduction in presynaptic VGCCs leads to a decrease in calcium influx during a presynaptic action potential, which decreases chemical neurotransmission, leading to a debilitating neuromuscular weakness (Lambert et al., 1956; Smith et al. 1995; Titulaer et al., 2011b; Tarr et al., 2015). Currently, the most common symptomatic treatment option for LEMS is a potassium channel blocker, 3,4-diaminopyridine (3,4-DAP). By blocking potassium channels in the presynaptic nerve terminal, the action potential is broadened and the duration of membrane depolarization is increased. This prolonged period of depolarization activates a greater number of presynaptic VGCCs and increases calcium influx into the nerve terminal (Verschuuren et al., 2006; Oh et al., 2009; Wirtz et al., 2009). However, there are dose-limiting side effects with 3,4-DAP, including paresthesia, gastric symptoms, insomnia, and less commonly, seizures (Verschuuren et al., 2006; Oh et al., 2009; Titulaer et al., 2011a). As such, the doses of 3,4-DAP that LEMS patients are typically prescribed may only lead to a modest relief of symptoms, and most patients continue to experience significant impairment in their activities of daily life (Sedehizadeh et al., 2012).

An alternative to using potassium channel blockers to treat LEMS and other neuromuscular weakness disorders of presynaptic origin is to target directly the VGCCs that remain in the terminal (Tarr et al., 2015). The compound, (R)-roscovitine was originally developed as a cyclin-dependent kinase (cdk) inhibitor and is currently in clinical trials for cancer and cystic fibrosis (Meijer et al., 1997; Meijer and Raymond, 2003; Meijer et al., 2016; Roskoski, 2016; Mottier and Rault, 2016). (R)-Roscovitine was also shown to be a VGCC gating modifier and agonist (Yan et al., 2002; Buraei et al., 2005; Cho and Meriney, 2006; DeStefino et al., 2010). A modified analog of (R)-roscovitine acting with greater potency and efficacy on calcium channels, coupled with reduced activity as a kinase inhibitor, would clearly be therapeutically beneficial for the treatment of synaptic weakness. Previously, we reported the development of (R)-roscovitine analogs with selective VGCC activity that prolonged the mean open time of Cav2 VGCCs while having a reduced effect on cdks (Liang et al., 2012; Tarr et al., 2013). These agents hold promise for therapeutic development because they have been shown to reverse the neuromuscular weakness characteristic of LEMS (Tarr et al., 2013; 2014).

Herein, we further explore (R)-roscovitine analogs with the goal of increasing our understanding of the structure-activity relationship (SAR) of these compounds. We identify a specific biological activity profile that is useful in the evaluation of novel analogs for potential use as therapeutics for neuromuscular weakness. Specifically, we identify structural features of (R)-roscovitine derivatives that convey improved activity at VGCCs. These novel compounds are characterized using patch clamp recordings of VGCC current (to measure gating modifier effects), kinase assays, cell survival assays (as an additional indirect measure of cdk activity), and intracellular recordings from neuromuscular synapses (to assess biological efficacy).

2. METHODS

2.1. Development of novel (R)-roscovitine analogs

Our primary design strategy for a new series of analogs was to modify the number and the position of nitrogen atoms in the imidazolopyrimidine core of GV-58 (Fig. 1). Specifically, we chose to retain the methylthiophene substituent since it exhibited superior activity in our earlier SAR studies (Liang et al., 2012), and only perform minor modifications of the n-propyl and the (R)-2-aminobutan-1-ol side chains to create the pyrazolopyrimidines MF-06, MF-17, MF-21, MF-27, MF-59, and MF-71, and the triazolopyrimidines KK-75, KK-76, and EG-02. Since the N-allylated KK-75 showed an attractive increase in the speed of action vs the N-propylated KK-76 (See section 3.3 below), we also generated the corresponding N-allylated imidazolopyrimidine analog of GV-58, KK-20. These modified heterocycles provide subtle but biologically relevant modifications in the electron density of the aromatic ring system and vary the distance between side chain elements and core heteroatom lone pairs. We hypothesized that these modifications would induce changes in potency and efficacy as well as calcium channel dynamics.

Fig. 1.

Strategy for expansion of GV-58. A. Zones that define areas of focus for development of novel analogs based on the parent molecule (R)-roscovitine. B. Expansion of the GV-58 calcium channel agonist pharmacophore (blue) within zone 1 to the MF-pyrazolopyrimidines (red) and the KK/EG-triazolopyrimidines (green), including substitutions within zones 3 and 4.

Full details of the synthesis of all intermediates and analogs is presented as Supplemental Information.

2.2. Cell Lines

We tested Cav2.1 VGCC agonist gating modification in transiently transfected HEK293 cells. These cells were maintained in Dulbecco’s modified Eagle’s medium (DMEM) with 10% fetal bovine serum at 37°C and 5% CO₂, and split twice per week. Upon splitting, small aliquots of cells were plated onto 35 mm tissue culture dishes for transfection and subsequent electrophysiological recording. After plating, cells were transiently transfected using FuGENE 6 (Promega Corp., Madison, WI, USA) with CaV 2.1 alpha1 subunits, CaVβ3, CaVα2δ1, and GFP at a DNA ratio of 1:1:1:1 (Addgene, Inc., Cambridge, MA, USA). Recordings of current through calcium channels were made from isolated fluorescently identified cells 24–72 h after transfection.

To evaluate effects of analogs on cell survival, SH-SY5Y cells (ATCC, Manassas, VA) were used in a cell proliferation assay. SH-SY5Y cells were maintained in a 1:1 mixture of ATCC-formulated Eagle’s Minimum Essential Medium (EMEM), and F12 Medium with fetal bovine serum at a final concentration of 10% at 37°C and 5% CO₂, and were split when they reached 90% confluence.

2.3. Whole-Cell Perforated Patch Clamp Recordings

Whole-cell perforated patch clamp recordings were performed on HEK293 cells transfected with CaV2.1 as previously described (Tarr et al., 2013; 2014). All recordings were performed at room temperature (20–22 °C). Cells were bathed in a saline containing 130 mM choline chloride, 10 mM Hepes, 10 mM TEA-Cl, 5 mM BaCl₂, and 1 mM MgCl₂ at pH 7.4. The pipette solution contained 70 mM CH₃C_SO₃S, 60 mM CsCl, 10 mM Hepes, and 1 mM MgCl₂ at pH 7.4. Patch pipettes were fabricated from borosilicate glass and pulled to a resistance of ~1–4 MΩ. Before each experiment, 3 mg amphotericin-B was dissolved in 50 μL DMSO. Approximately every hour, 10 μL of this stock amphotericin solution was mixed with 0.5 mL filtered pipette solution. Pipettes were filled in a two step process. The tip was dipped into a droplet of filtered pipette solution for 1–3 sec, and then the remainder of the pipette was back-filled with the amphotericin-B/pipette solution mixture using a syringe and a 34 G quartz needle (MicroFil MF34G, World Precision Instruments, Sarasota, FL). This pipette was then used to make a GΩ seal with a fluorescent cell and given time for the amphotericin-B to perforate the patch of membrane isolated under the pipet in order to gain electrical access to the whole cell (5–10 min). Access resistances ranged from 10–20 MΩ and were compensated by 80–85%. All chemicals were obtained from Sigma-Aldrich Corp. (St. Louis, MO, USA) unless otherwise noted.

Voltage clamp of cells was controlled by an Axopatch 200B amplifier driven by Clampex 9 software (Molecular Devices, Sunnyvale, CA). Data were filtered at 5 kHz and digitized at 10 or 50 kHz. Data were analyzed using Clampfit 10 software (Molecular Devices). Capacitive transients and passive membrane responses to the voltage steps were subtracted, and a liquid junction potential of −11.7 mV was corrected before each recording. Current through calcium channels was activated by depolarizing steps from a holding potential of −100 mV to −50 through +90mV in 10 mV steps for 20–100 msec before returning to −100 mV for the generation of IV and voltage-dependence of activation curves, and from −100 mV to +20 or +60 mV for 1–200 msec before returning to −60 mV for evaluation of GV-58 and other analog actions on tail current. These depolarizing steps were given every 5–30 seconds to allow VGCCs to recover between stimuli. All analogs were dissolved in DMSO at 50 mM to create a stock solution, aliquoted, and stored at − 20 °C before use. On the day of use, stock solutions were diluted into warmed extracellular saline at the indicated concentrations (1–100 μM) for application to cells or synapses. In all final salines, DMSO was kept at or below 0.1% (v/v), which was shown to have no effects on calcium channel current in patch clamp recordings (Tarr et al., 2013). For most experiments, drug solutions were delivered directly to a cell under study using a pressurized borosilicate glass pipette lowered over the cell during the recording. For experiments in which analogs were applied rapidly to cells during patch clamp recordings, rapid solution exchange was achieved using an in-house-fabricated fast perfusion system (Qian et al., 2002) connected to gravity-fed reservoirs. For these experiments, the solution exchange time constant was determined to be 27.4 ± 6.6 msec (mean ± SD; Glasgow et al., 2017)

Peak currents during the voltage step, and tail currents at the conclusion of the voltage step, were recorded from the whole cell both before and during drug application to the cell. The integral of the tail current was measured and normalized to its peak to correct for any inactivation of calcium channels during the experiment. Efficacy of novel analogs was defined as increases in the tail current integral above control.

The time course of current activation was fit by a single exponential. Deactivation of tail currents was fit by a double exponential. Analogs tested here prolonged the slow time constant for deactivation without affecting the fast time constant as has previously been reported for (R)-roscovitine (DeStefino et al., 2010). In some cases there was no fast time constant that could be measured when analogs were highly effective (after very long depolarizations at high analog concentration). In these cases, the slow time constant for decay was measured starting 1 msec after the peak of the tail current.

2.4. Kinase Activity and Cell Survival Assay

To evaluate the effect of selected analogs on kinase activity, the kinaseseeker assay was performed by Luceome Biotechnologies, LLC (Tuscon, AZ) as previously described (Jester et al., 2010; 2012). This is a binding assay in which displacement of a labeled probe by a candidate kinase inhibitor is measure in a luminescence assay. In this cell-free assay using rabbit reticulocyte lysates, selected calcium channel gating modifier analogs (at 2 μM),or DMSO (control), were incubated for 30 minutes at room temperature, followed by 1 hour in the presence of a kinase specific probe. Luminescence was measured using a luminometer. Values were expressed as percentage of kinase activity remaining at the conclusion of the assay.

A cell survival assay was performed to confirm physiologically relevant cdk antagonist effects of (R)-roscovitine and selected analogs. For this assay, SH-SY5Y cells (a human neuroblastoma cell line) were chosen as they have been used previously to study effects of cdk inhibition on cell survival (Ribas & Boix, 2004; Ribas et al. 2006). Cells in each well of a 96-well plate were incubated in culture medium at 37° C and 5% CO₂ for 24 h, and then exposed to (R)-roscovitine or one of its analogs at a concentration of 0.1, 0.5, 1, 10, 50, or 100 μM for an additional 24 h. After this overnight incubation, 20 μL of MTS tetrazolium-based reagent was added to each well and incubated for 4 h according to kit instructions (CellTiter 96 Kit; Promega Corp). Absorbance was read at 490 nm in each well using an Infinite 200 PRO microplate reader (Tecan Trading AG, Mannedorf, Switzerland) to determine the percentage of surviving cells (normalized to vehicle; 0.2% DMSO). Wells in which no cells were added were used to determine background absorbance, which was subtracted from each reading.

2.5. Intracellular recordings from model neuromuscular synapses

Intracellular recordings of endplate potentials (EPPs) and miniature endplate potentials (mEPPs) were recorded and analyzed using the pClamp software suite (Molecular Devices) from the mouse epitrochleoanconeus (ETA) muscle as previously described (Tarr et al., 2013). Briefly, adult female mice (outbred Swiss Webster) were killed by CO₂ inhalation in accordance with University of Pittsburgh Institutional Animal Care and Use Committee guidelines. Dissected nerve-muscle preparations were pinned to the bottom of a Sylgard-coated 35 mm petri dish and bathed in normal mammalian saline (150 mM NaCl, 5 mM KCl, 11 mM glucose, 10 mM Hepes, 1 mM MgCl₂ and 2 mM CaCl₂, pH 7.3–7.4). Intracelluluar recordings from peri-synaptic regions of individual muscle fibers were performed using 40–60 MΩ borosilicate glass electrodes filled with 3 M potassium acetate. The muscle nerve was stimulated at 0.2 Hz using a section electrode at 10 times the intensity required to elicit muscle contractions, which were subsequently blocked by exposure to 1 μM ω-conotoxin GIIIB (Alomone Labs Ltd, Jerusalem, Israel). To generate a pharmacological model of presynaptic neuromuscular weakness, the nerve-muscle preparation was exposed to 50–100 nM ω-agatoxin IVA (to partially block voltage-gated calcium channels) for 20–30 minutes, or until EPPs were between 2–10 mV. These weakened synapses were then used to test the effects of a 30 minute exposure to vehicle (0.1% DMSO) or 50 μM GV-58, KK-75, KK-20, or MF-06. In all cases, the same synapses were recorded before and after exposure to GV-58, KK-75, KK-20, or MF-06 so that changes in transmitter release could be determined within each synapse. Control recordings were made before drug exposure with 0.1% DMSO in the bath. For each drug condition, we recorded from 24–35 synapses taken from 4–6 mice. Since transmitter release varied significantly between individual synapses, but not between animals, the number of experiments (n) was defined as the number of synapses. The magnitude of transmitter released was calculated in each paired recording as quantal content (QC). For this determination, the mean amplitude of 10–20 EPPs were first corrected for non-linear summation (McLachlan and Martin, 1981), and this corrected mean EPP amplitude was then divided by the mean mEPP amplitude (average of >100 events).

2.6. Statistical Analysis

Statistical analyses were performed using Origin 7 (OriginLab Corp., Northampton, MA, USA) or GraphPad Prism version 5.0 (GraphPad Software, Inc., La Jolla, CA, USA), with a significance level of P < 0.05 for all statistical tests. One sample comparisons or comparisons between two groups employed the t-test, while comparisons with multiple groups used a one-way analysis of variance with Tukey’s post-hoc test. Data are given as mean ± SEM unless otherwise noted. For patch clamp tests of agonist activity, each drug concentration or voltage step was tested in 5–12 different cells. For the cell survival assay, each concentration of drug was measured in ≥ 6 wells. Curves for concentration-response effects on current through calcium channels and cell proliferation were fitted with the Hill equation: y = Vmax*xⁿ/(kⁿ+xⁿ), where x is the concentration of drug, y is the effect of the drug on the tail current integral (normalized to control), and n is the Hill coefficient.

3. RESULTS

3.1. Agonist gating modification at Cav2.1 VGCCs as a screen for novel analogs

(R)-Roscovitine is a known cdk inhibitor that also displays stereoselective agonist activity at Cav2 VGCCs (Yan et al., 2002; Buraei et al., 2005). While (S)-roscovitine is only 2-fold less potent as a cdk inhibitor than (R)-roscovitine (De Azevedo et al., 1997), (S)-roscovitine has no VGCC agonist activity unless the concentration is raised above 500 μM (a 20-fold lower affinity than (R)-roscovitine; Buraei and Elmslie, 2008). Previously, we reported the development of GV-58, an analog of (R)-roscovitine with a 4-fold higher potency and efficacy for Cav2 VGCCs as a calcium channel agonist gating modifier, and with a greater than 20-fold reduction in cdk activity than the parent molecule (Liang et al., 2012; Tarr et al., 2013).

We defined 4 zones of (R)-roscovitine for SAR modifications (Fig. 1A). Based on our previous experience in the development of GV-58, we retained the thiophene moiety in zone 2 that was shown to be critical for Cav2.1 VGCC agonist activity, and in this study focused on modifications to the other 3 zones (Fig. 1B). We began our screen of new analogs by testing Cav2.1 VGCC agonist activity at 50 μM (Fig. 2). Our first modification was to zone 1, replacing the purine with a pyrazolo[1,5-a]-1,3,5-triazine. Previous work in our lab has shown that these compounds retain VGCC agonist activity (unpublished, Liang et al., 2012). The analog MF-17 retains the zone 2, 3, and 4 substituents of GV-58 and showed similar efficacy to the parent compound. Since this zone 1 modification was well tolerated, we then investigated a small number of zone 3 modifications while retaining the pyrazolo[1,5-a]-1,3,5-triazine core. Replacing the ethyl with a trifluoromethyl group in MF-59 abolished the improvements gained with GV-58, such that MF-59 showed similar activity to (R)-roscovitine. The gem-dimethyl analog MF-21 performed even less effectively, leading to a greater than six-fold reduction in observed tail current integrals relative to GV-58, while the ester derivative MF-27 showed a complete loss of activity. We also confirmed that (S)-GV-58 is less effective than (R)-GV-58, although the difference in efficacy between these enantiomers was not as great as observed in the parent molecule, roscovitine. Taken together, these results are consistent with prior literature results that suggest agonist activity is very sensitive to zone 3 modifications (Buraei and Elmslie, 2008).

Fig. 2.

VGCC gating modifications of novel analogs. A. Structures of (R)-roscovitine and analogs developed as agonist gating modifiers. Dotted lines circle structural modifications as compared to (R)-roscovitine. Structures in black did not have VGCC modifying properties that warranted further study. Colored structures were studied further below and data employ the same color coding. B. Plot of total current integral during the deactivation phase of current, normalized to the integral of control (unmodified) current, for all 10 analogs shown in A. These currents were activated by 200 msec voltage steps from −100 mV to +60 mV, with a return to −60 mV for the measurement of the tail currents, as shown in C. C. Sample VGCC currents recorded before (black) and after a 15–30 second exposure to each analog (color coded as shown in A). All of the six colored analogs slowed deactivation kinetics (D) to a greater degree than observed with (R)-roscovitine (light blue). D. Plot of the slow deactivation time constant (tau in ms) for analog-modified calcium channel tail currents.

Two zone 4 modifications of the pyrazolo-triazine scaffold were investigated further (Fig. 2). Unfortunately, aqueous solubility limitations of the cyclopropylmethylene compound MF-71 (Fig. 1) prevented us from analyzing the biological properties of this analog. Replacing the n-propyl with an iso-propyl group in MF-06 led to the largest observed increase in the tail current integral at ~1.5 times that of GV-58. The strong effect of MF-06 (Fig. 2B & C) was a function of a significantly prolonged tail current decay time constant (Fig. 2D) which was a >2-fold increase over GV-58 (21.0 ± 2.45 msec and 9.16 ± 0.706 msec respectively). This predicts a significantly increased mean open time for modified single VGCCs after exposure to MF-06.

We also investigated three [1,2,3]triazolo[4,5-d]pyrimidine derivatives which incorporate an additional nitrogen atom in zone 1 (Fig. 1). KK-76, the core modified analog of GV-58, as well as KK-75, the zone 4 unsaturated derivative, both showed slightly reduced efficacy compared to the parent compound (Fig. 2), but possess other interesting properties (see Section 3.3 below). EG-02 was prepared in order to assess the effect of branching in the zone 4 side chain, but since this analog was less effective than GV-58 we did not study it further. Finally, the attractive kinetic properties of KK-75 also inspired us to prepare a zone 4 unsaturated analog of GV-58, the imidazolopyrimidine KK-20. KK-20 had an efficacy that was similar to GV-58 (Fig 2), and had interesting kinetic properties (see Section 3.3 below).

3.2. Concentration-response relationships

To evaluate the concentration-response relationship for the five most effective analogs (MF-06, MF-17, KK-75, KK-76, and KK-20) as compared with GV-58, we tested agonist effects on VGCC current using at least 4 different concentrations ranging from 1 to 100 μM (Fig. 3). For some compounds, aqueous solubility limitations prevented us from testing at higher concentrations. Overall, the EC₅₀ values for MF-17, KK-75, KK-76, and KK-20 were not significantly different than that of GV-58 (ranging from 7.04–11.42 μM; see Table 1). However, the EC₅₀ value for MF-06 agonist activity on Cav2.1 VGCCs was 2.49 μM, which was significantly lower than GV-58 (7.14 μM). All of our novel analogs had EC₅₀ values that were significantly lower than (R)-roscovitine (27.58 μM; Liang et al., 2012).

Fig. 3.

Concentration-response relationships for six of our most interesting analogs (color coded as in Fig. 2). Currents were activated, measured, and analyzed as described in Fig. 2. Fitted curves do not extend beyond the concentrations in which DMSO stock solutions were soluble in recording saline.

Table 1.

EC50 values obtained by fitting concentration-response curves for analog effects on the integral of calcium channel tail currents.

Drug	EC50 (μM)
R-ros	27.58
GV-58	7.14
MF-06	2.49
MF-17	11.42
KK-75	10.28
KK-76	7.11
KK-20	7.04

3.3. Speed of state-dependent gating modification

Previous studies of (R)-roscovitine (Buraei et al., 2005) and GV-58 (Tarr et al., 2014) have reported that both compounds modulate the gating kinetics of calcium channels only when they are in the open state. This state-dependent gating modification was tested using depolarizing steps to +20 mV of varying duration (1–100 msec) in the continuous presence of our novel analogs (all at 50 μM; Fig. 4A & B). This experiment probed how quickly novel analogs modify the gating of open channels by testing the proportion of current through calcium channels that is modified when VGCCs are in an open state for very brief periods of time. Characterizing this property is critical to predicting which of these novel analogs might be of therapeutic interest, since neuronal action potentials only open VGCCs for very brief periods of time (1–2 msec). For easier comparison of state-dependent gating, data were normalized to the largest tail current integral value. As expected of these compounds, all analogs showed a decrease in agonist effect as the duration of the depolarizing step was reduced (Fig. 4C).

Fig. 4.

Calcium channel state-dependent effects of novel analogs. A. Sample tail currents evoked by depolarizing steps (from −100 mV to +20 mV and returning to −60 mV) of varying duration (1, 2, and 100 msec) in control (black) and in the presence of MF-06 (color coded as shown in Fig. 2). B. Sample currents evoked by depolarizing steps as decribed in A of varying duration (1, 2, and 100 msec) in control (black) and in the presence of KK-75 (color coded as shown in Fig. 2). C. Plot of the impact of depolarizing step duration on the magnitude of agonist gating modifier effects on tail current integral. Data are fit to the Boltzmann equation. As depolarization time is increased, gating modifiers have a stronger effect. However, the specific relationship is shifted depending on the analog tested, with some having little effect during brief 1–2 msec depolarizations (e.g. MF-17 and MF-06), and others demonstrating ~25–45% of maximal effects during such brief depolarizations (e.g. KK-75). Data are color coded for each analog as in Fig. 2. All analog effects on tail current are normalized to their individual maximums (tail current following a 100 msec step depolarization).

These data provide evidence that the pyrazolo-triazine zone 1 analogs MF-06 and MF-17 were too slow to significantly modify VGCC gating when depolarizations were less than 2 msec in duration. In contrast, triazolo-pyrimidine analogs KK-76 and KK-75 retained a similar kinetic profile to GV-58, with the zone 4 unsaturated derivative (KK-75) showing a slight increase in the speed of action. Further, the N-allylated imidazolopyrimidine analog KK-20, which also contains an unsaturated zone 4 side chain, also showed increased speed of action, particularly for depolarizations longer than 2 msec in duration (Fig. 4C). Therefore, in the continuous presence of our novel analogs, when the calcium channel was triggered to move into the open state by depolarization, the fastest of these analogs (GV-58, KK-75, KK-76, and KK-20) displayed 50% of maximal effects within 2–4 msec, while the slower analogs (MF-17 and MF-06) displayed 50% of maximal effects within 8–11 msec (Fig. 4C).

3.4. Speed of Action During Fast Perfusion of Analogs onto Cells

While the data presented in section 3.3 above document the speed of open state-dependent gating modification in the continuous presence of analogs, we were also interested in how fast these analogs could modify the calcium channel when perfused onto the cell. To examine this property, we chose to study KK-20 as a representative analog among those that showed fast state-dependent action (Fig. 4C). For these experiments we employed a fast perfusion system (Qian et al., 2002; Glasgow et al., 2017). Our initial goal was to use this fast perfusion system to apply KK-20 for a brief period of time (50–100 ms) during a depolarizing step (to +10 mV) and measure the on and off rates of effects on current during the depolarizing step. During our attempts to perform these experiments it became clear that the off rate was so slow that very long depolarizing steps would have been required (>500 ms), which would have resulted in nearly complete voltage-dependent current inactivation. Instead, we examined how quickly 50 μM KK-20 could modify calcium channel tail current when applied at different time points after the start of a 200 msec depolarizing step (Fig. 5A & B). For this experiment, channels were already in the open state at the time of KK-20 application. Compared to the maximal effects of 50 μM KK-20 (as determined by gating modification in the contiuous presence of KK-20 for at least 1 second before and during the depolarizing step), application only during the depolarizing step was not very effective. For example, when 50 μM KK-20 exposure was initiated during the depolarizing voltage step, about 150 msec before tail current measurement, KK-20 only increased tail current by 25–30% of maximum (Fig. 5B). This suggests that either KK-20 is not very efficacious when applied at +10 mV, the EC₅₀ for KK-20 increases with depolarization, or the kinetics of KK-20 action are so slow when applied at +10 mV that more than 150 msec is required for it to fully bind to calcium channels. Further experiments will be needed to address these issues.

Fig. 5.

Effects of analogs following rapid solution exchange. A. Rapid application of 50 μM KK-20 during a 200 msec depolarizing step to +10 mV followed by a return to −100 mV to measure tail currents. Top traces show the time course of KK-20 application (Drug Delivery). Middle trace shows the depolarizing voltage step. Bottom trace on right shows calcium channel tail currents. Colors of lines indicating onset of KK-20 application (top traces) match colors of tail currents (bottom-right traces). The tail current shown in the dotted line represents the maximum tail current elicited by the application of KK-20 for 1 sec before the the tail current measurement at the conclusion of the voltage step. B. Plot of the effect of KK-20 on tail current integral. Data are normalized to the maximum tail current integral shown with the dotted line in A as a function of the length of drug application during the voltage step used to activate current. C. Rapid application and removal of 50 μM KK-20 before the onset of the voltage step to activate current. As this 100 msec drug delivery pulse is moved further in advance of the voltage step to activate current (as indicated by color coded times), the effects of KK-20 are reduced. Tail currents corresponding to intervals between the conclusion of the drug delivery and the end of the voltage step are color coded as indicated by the delay interval label. D. Plot of the effect of KK-20 on tail current integral (normalized to the maximum as described above) as the delay interval between drug delivery and the voltage step is increased. Data points are fit with a single exponential (red line) with tau = 0.99 seconds.

These results led us to examine how effectively KK-20 could modify calcium channels when applied at different times before depolarization. We found that a 100 msec application of 50 μM KK-20, applied 100 msec before a depolarizing voltage step (while the cell was held at −100 mV), was very effective (more than 70% of maximum) at modifiying deactivation kinetics of the current (Fig. 5 C & D). We next investigated the time course of KK-20 unbinding at −100 mV. We found that as the time between the 100 ms application of 50 μM KK-20 and the depolarizing voltage step increased, the calcium channel tail current integral decreased. Based on a single-exponential fit to the time course of tail current decay, we estimate that the time constant of KK-20 unbinding at −100 mV is 0.99 s (Fig. 5D). Taken together, the data reported in figure 5 support the hypothesis that KK-20 binding to the calcium channel is more effective at modifying tail current when the channel is in the closed state (holding potential of −100 mV). The mechanisms that underlie these observations will require further study.

3.5. Effects of Novel Analogs on the Activation of Calcium Channels

To examine the effects of our novel analogs on the voltage dependence of activation, we activated calcium channels with a series of 20 msec voltage steps (to −50 through +90 mV) from a holding potential of −100 mV in control saline and in the continued presence of 50 μM GV-58, MF-06, KK-75 or KK-20. These were paried recordings in the same cells before and during drug application (n = 6–8). When plotting peak current during the step at each depolarized voltage, the IV plots showed about a 10 mV shift to the left after analog application (Fig. 6A,C, E & G). However, when peak tail current was plotted against the voltage step used to trigger calcium current, the resulting voltage-dependence of activation curves did not show any shift (Fig. 6B, D, F & H). We suspected that the discord between the IV and voltage-dependence of activation plots was due to the presence of a contaminating rectifying outward current in HEK293 cells that was present during the voltage steps, despite the presence of intracellular CsCl and extracellular 10 mM TEA to block potassium channels. Such currents have been previously described (Jiang et al., 2002). Our preliminary investigation into this contaminating current reveals that it was of variable size (0 to 230 pA at +60 mV; n = 7 cells), and was increased by exposure to 50 μM GV-58 in 4 of 7 cells studied. The GV-58-mediated increase in rectifying current amplitude averaged 70 ± 10% (mean ± SEM; n = 7) across all voltages tested (20 msec steps from −100 mV to −50 to +90 mV). This contaminating current did not have a significant contribution during the calcium channel tail current measurements, and as such, only contributed significantly during peak calcium channel current measurements used to generate IV curves. Therefore, we conclude that our novel analogs do not shift the voltage-dependence of activation for calcium channels.

Fig. 6.

Effects of novel analogs on current-voltage relationships (IV curves) and the voltage dependence of current activation. For these experiments, currents were activated from a holding potential of −100 mV by 20 msec depolarizing steps to −50 mV through +80 mV in 10 mV increments, followed by a return to −100 mV. IV curves (A, C, E & G) were generated by plotting peak current through calcium channels during the voltage step against the magnitude of the voltage step. Voltage-dependence of activation curves (B, D, F & H) were generated by plotting peak tail current at −100 mV after the step against the magnitude of the voltage step used to activate current. Control curves are shown in black, and effects of 50 μM GV-58 (A & B), MF-06 (C & D), KK-75 (E & F), and KK-20 (G & H) are shown in red.

To explore the effects of analogs during the activation of calcium channels, we measured the kinetics of calcium channel activation following a step from −100 mV to +10 mV for 100 msec before and during application of analogs. As expected, because our novel analogs take several milliseconds to modify channels following a depolarizing voltage step (see Fig. 4), and these gating modifier analogs increase mean open time of the channel (DeStefino et al., 2010; S.D. Meriney, unpublished observations), the activation of calcium current increased during the first 5–10 msec of the voltage step to a level that was greater than observed in control conditions (Fig. 7A). When the activation phase of the current was fit with a single exponential, there was a tendancy for analog-modified current to have slower activation kinetics (Fig. 7B). In the presence of KK-75 and KK-20, this slowing in activation kinetics was statistically significant.

Fig. 7.

Effects of novel analogs on activation kinetics of current through calcium channels. A. Sample currents evoked by voltage steps from a holding potential of −100 mV to +10 mV for 100 msec, followed by a return to −100 mV. The black trace is a control current, and the red trace is the analog-modified current after exposure to 50 μM KK-20 for 5 seconds before the onset of the voltage step. B. Plot of activation kinetics (single exponential tau in msec; mean +/− SEM) before (black bars) and after (red bars) exposure to GV-58, KK-20, KK-75, and MF-06 as indicated (n = 6–8 cells). * significantly different from control measurement before drug application (p < 0.05; paired t-test).

Measuring effects of our analogs on inactivation of current through calcium channels is complicated by the fact that these gating modifiers have slow action that increases the proportion of channels gating with longer mean open time (see Fig. 4) after the initiation of a voltage depolarization step. As such, the dynamic change in the mean open time of channels during the voltage step makes measurement of inactivation inaccurate (see Fig. 7A).

3.6. Kinase Activity and Cell Survival

Because (R)-roscovitine is a potent inhibitor of cell cycle regulating cdks, we tested the activity of our novel analogs in a kinase inhibition assay (kinaseseeker, Luceome Biotechnologies, LLC, Tuscon, AZ). For this assay we measured kinase inhibition of cdk2, cdk5, calcium calmodulin-dependent kinase 2a (CAMK2a), Ephrin 2B receptor tyrosine kinase (EPH2B), and ribosomal S6 kinase 1 (RSK1) following application of selected analogs (at 2 μM). None of the analogs tested reduced EPHB2 activity by more than 0–10%. Similarly, CAMK2b and RSK1 activity were little affected by these analogs (5–25% effects). Effects on cdks were analog dependent. While GV-58, KK-75, MF-17, and KK-20 only reduced cdk5 activity by 10–20%, MF-06 reduced cdk5 activity by about 40% (Fig. 8A). With regard to cdk2, GV-58, KK-75, and KK-20 had minimal effects (reduction in activity of 0–15%). In contrast, MF-17 reduced cdk2 activity by about 25%, and MF-06 reduced cdk2 activity by 75% (Fig. 8A).

Fig. 8.

Effects of analogs on kinase activity and survival in a culture of dividing cells. A. Effects of five analogs (GV-58, MF-17, MF-06, KK-75, and KK-20) at 2 μM on kinase activity (cdk2, cdk5, CAMK2a, EPHB2, and RSK1) as color coded and indicated in the legend. 100% represents full, unaffected kinase activity. B. Cell survival of SH-SY5Y cells in the presence of six different calcium channel agonist gating modifiers at varying concentrations. All data are normalized to cell survival in the presence of vehicle (0.2% DMSO) and subtracted from background readings in the absence of cells. Different symbols represent data from each analog (as noted in the keys). Analogs are grouped into pairs based on their effects on cells survival. (R)-roscovitine and MF-06 had similar results and a red line was fit to data from both. KK-75 and KK-76 had similar results and a green line was fit to data from both. GV-58 and MF-17 had similar results and at blue line was fit to data from both. The dotted line box highlights the results at 50 μM in which GV-58, MF-17, KK-75, and KK-76 had no significant effects on cell survival. Red asterisk indicates significantly different from 1.0 (no change) using one-sample t-test.

In addition to activity in the kinase assay reported above, we also tested our analogs for cdk-dependent cellular effects by measuring the survival of rapidly dividing cells in vitro. The proportion of cells that survived 24 hours after exposure to varying concentrations of (R)-roscovitine, GV-58, MF-06, MF-17, KK-75, and KK-76 were measured using the MTS-based cell-survival assay and plotted in Figure 8B. Based on the results, three pairs of compounds with similar cell survival profiles emerged. GV-58 and MF-17 showed little or no effect on cell survival, even up to 100 μM: MF-17 did not show a significant decrease in cell survival at any concentration, and GV-58 only decreased cell survival by 8% at a concentration of 100 μM (Tarr et al., 2014). KK-75 and KK-76 had no effects on cell survival up to 50 μM, but leading to a ca. 50% reduction in cell survival at 100 μM. In contrast, (R)-roscovitine and MF-06 showed a significant decrease in cell survival at concentrations of 10 μM or above (Fig. 8B). Differences in results between this cell survival assay and the kinase assay reported above may be due to differences in ATP concentrations in the two assays, as roscovitine and its analogs compete with ATP for binding to cdks.

3.7. Modification of transmitter release

To evaluate whether the most promising of our novel analogs could increase transmitter release at weakened neuromuscular junctions, we used a pharmacological model of neuromuscular weakness (partial block of presynaptic VGCCs using ω-agatoxin IVA) and measured effects on the magnitude of transmitter release (as estimated using quantal content – the number of synaptic vesicles released with each presynaptic action potential). These experiments permitted us to compare the impact on transmitter release of analogs with different potencies, efficacies, and speed of action during a natural presynaptic action potential depolarization. In this analysis, we chose to compare MF-06 (the analog with the strongest potency (Fig. 3) and efficacy (Fig. 2), but slowest speed of action (Fig. 4)), KK-75 (the analog with the fastest speed of action (Fig. 4), but weaker efficacy(Fig. 2)), and KK-20 (the analog with relatively fast speed of action and strong potency) to our previously developed molecule GV-58 (see Figs. 2, 3 & 4). While GV-58 significantly increased transmitter release (p < 0.05) by more than two-fold, neither MF-06 nor KK-75 were as effective (Fig. 9). KK-75 showed a tendency to increase quantal content, but the effects were not significant (p > 0.13). MF-06 had a weaker effect on synapses that was not significant (p = 0.58). Interestingly, KK-20 significantly increased transmitter release (Fig. 9; p < 0.05; one-way analysis of variance with Tukey’s post hoc test).

Fig. 9.

Effects of selected analogs on transmitter release at the mouse neuromuscular junction. A. Plot of the magnitude of analog effects on transmitter release evoked by single action potentials delivered to the muscle nerve (measured as quantal content). While the vehicle (0.1% DMSO) had no significant effect on transmitter release (n = 28 synapses), 50 μM GV-58 more than doubled transmitter release (n = 32 synapses), 50 μM KK-20 increased transmitter release by about two-fold (n = 23 synapses), KK-75 had a tendancy to increase transmitter release but the effect was not significant (n = 24 synapses), and 50 μM MF-06 had no significant effect (n = 29 synapses). * significantly different from DMSO, one-way analysis of variance with Tukey’s posthoc test. B. Representative sample EPPs recorded at a single synapse before (black trace) and after exposure to each analog as listed (red trace). Traces shown are an average of 10–20 sweeps elicited by single action potentials stimulated at 0.2 Hz.

4. DISCUSSION

The present research extends our development of (R)-roscovitine derivatives with agonist activity at Cav2 VGCCs (Liang et al., 2012) that have therapeutic potential to treat neuromuscular weakness (Tarr et al., 2013; 2014). In this study, we have explored the SAR of ten novel analogs for agonist activity at Cav2 VGCCs. Furthermore, we have characterized some of the biophysical properties of calcium channels that are modified by the most interesting of these novel analogs.

4.1. Cdk activity

The parent molecule (R)-roscovitine was developed as a cdk inhibitor and entered clinical trials as a treatment for cancer and other disorders (Meijer et al., 1997; Meijer and Raymond, 2003; Meijer et al., 2016; Roskoski, 2016). A major goal of our analog development strategy is to significantly reduce cdk activity to create more selective Cav2 gating modifiers. Consistent with our previous SAR studies (Liang et al., 2012; Tarr et al., 2013), the substitution of an iso-propyl (e.g. MF-06) with an n-propyl (e.g. MF-17) side group in zone 4 greatly reduced cdk activity as demonstrated by reduced effects in our kinase assay and on cell survival (Fig. 8), without eliminating Cav2 activity (Fig. 2). This particular change appears to reduce cdk activity independently of differences in other zones. This effect is likely due to the modification of a branched side chain that has been shown to be important for interaction within a hydrophobic pocket in the ATP binding site of cdk (Hardcastle et al., 2002).

4.2. Efficacy of drug action in vitro

The efficacy of action on Cav2 VGCCs is thought to reflect differences in modulator-induced stabilization of the open state of the VGCC, which results in lengthening of the mean open time of the channel (DeStefino et al., 2010). To explore the impact of the zone 1 heterocyclic core on the efficacy of action in vitro, we compared the degree to which the decay of VGCC current was lengthened and current integrals were increased (Fig. 2B and C). In comparing our novel analogs, those with a pyrazolopyrimidine core (MF-17 and MF-06) prolonged deactivation kinetics to the greatest degree and had the strongest efficacy (Fig. 2), while the imidazolopyrimidine core in GV-58 and in KK-20 had intermediate efficacy, and those with a triazolopyrimidine core (EG-02, KK-75 and KK-76) had weaker efficacy. Further, within the pyrazolopyrimidines (MF-06 and MF-17), the presence of an iso-propyl side chain in zone 4 (MF-06) resulted in higher efficacy (Fig. 2) and potency (Fig. 3) than when an n-propyl side chain was present (MF-17).

4.3. Effects of Analogs on Voltage-Dependent Activation of Calcium Channels

The absence of effects of our analogs on the voltage-dependence of activation as measured using tail current measurements (Fig. 6B, D, F & H) supports the conclusion that these analogs do not change this property of voltage-dependent calcium channels. The roughly 10 mV shift observed in IV plots of peak calcium current (Fig. 6A, C, E & G) is likely an artifact related to potential analog effects on a outward current native to HEK293 cells that is resistant to the potassium channel blockers included in our recording solutions (see Section 3.5). If our analogs do increase a voltage-dependent potassium current at synapses, this would be predicted to reduce transmitter release. However, since our analogs increase transmitter release, consistent with agonist effects on voltage-gated calcium channels, any potential enhancement of a potassium current may not be significant during brief action potential stimuli (for a variety of potential reasons, including the kinetics of such off-target action). With regard to off-target effects of our analogs on other types of channels, this is an interesting topic for those compounds that advance further in pre-clinical testing, or become popular tools for experimental study. However, this is a very large area of investigation that is beyond the scope of this report focused on the SAR of novel analogs at the voltage-gated calcium channel.

Consistent with their mechanism of action at voltage-gated calcium channels (to increase mean open time), our novel analogs do slow the activation kinetics of peak calcium current during a depolarizing voltage step (Fig. 7). This effect is likely caused by an increase in the proportion of calcium channels gating with longer mean open time after the voltage step is initiated since it can take about 10 msec for some analogs, and 100 msec for other analogs, to fully modify calcium current (see Fig. 4).

4.4. Speed of Analog action

There appear to be two steps contributing to the speed with which our novel gating modifiers can affect Cav2 calcium channels. First, the speed with which these modifiers can bind to the channel following initial exposure, and second, how fast bound drugs can modifiy gating to prolong mean open time after the channel opens. To address speed of binding we examined KK-20 as a representative compound using a fast perfusion system.

Because KK-20 is neutral at physiological pH (pKa1 = 5.53, pKa2 = −0.26), and not soluble in aqueous solutions, we consider it possible that it either binds in a hydrophobic pocket within the extracellular regions of the calcium channel, or partitions into the membrane before binding to the channel. If partitioning into the plasma membrane is required, this may be followed by access to the channel binding site through a fenestration in the channel protein, as has been described for other channels, particularly in evolutionarily related voltage-gated sodium channels (Hille, 1977; Payandeh et al., 2013; Leneus et al., 2014; Boiteux et al., 2014). Interestingly, our data (Fig. 5) lead to the hypothesis that binding of KK-20 to the calcium channel is more effective, or proceeds more quickly when the membrane is held at −100 mV than when it is depolarized to +10 mV. Therefore, it is possible that the binding site or membrane access to fenestrations in the channel are more available when the channel is in the closed configuration. Such state-dependent binding or fenestration access has been described for voltage-gated sodium channels (Hille, 1977; Kaczmarski and Corry, 2014). Details regarding the binding and action of our analogs at the voltage-gated calcium channel are likely to be complex and will require further investigation that is beyond the scope of this report.

Because our novel Cav2 agonist modifiers affect calcium channel gating only when channels are in the open state (Fig. 2 E-G; Buraei et al., 2005; Tarr et al., 2013), their kinetics of gating modificiation is a critical characteristic that predicts the potential for in vivo efficacy at neuromuscular synapses because presynaptic Cav2 VGCCs are normally activated by very brief (1–2 msec) action potentials. During these brief periods of activation, Cav2 channels usually only flicker open for less than 1 msec, providing a very short time window for modulator action. To explore the impact of zone 1 modifications specifically on this important property of our novel modulators, we left all other zones unchanged and varied the zone 1 core of the molecule (comparing GV-58 with KK-76 and MF-17). In this analysis, the parent GV-58 with an imidazolopyrimidine core appeared to act faster than KK-76 with a triazolopyrimidine core, while MF-17 with a pyrazolopyrimidine core had the slowest kinetics of action. Kinetics were also affected by zone 4 modifications, as the triazolopyrimidine core coupled with an alkene-containing side chain (KK-75) created an analog with the fastest speed of action over the first two msec of the depolarization (see Fig. 4C). This suggests that zone 4 modifications hold the potential to strongly improve binding kinetics and reduce cdk activity simultaneously and should be further investigated in the future.

4.5. Therapeutic potential

In developing novel analogs for potential use as therapeutics to treat neuromuscular weakness, the most promising compounds should act quickly (during brief presynaptic action potentials) and have strong potency and efficacy. In this context, GV-58 has a balance of these characteristics that creates a greater than twofold increase in transmitter release triggered by action potential activity. In our evaluation of the novel analogs presented here, we identified one compound with greater potency and efficacy than GV-58 (MF-06), but this analog had very slow kinetics of action. We also identified a novel analog with faster speed of action than GV-58 (KK-75), but this analog had lower efficacy. KK-20, a compound that combined a relatively fast speed of action and strong efficacy, significantly increased transmitter release to a level that was not significantly different from GV-58. Among the two novel analogs that did not have significant effects on transmitter release compared to the DMSO vehicle, KK-75′s fast speed of action was likely offset by a low efficacy, while MF-06’s strong efficacy was likely offset by a very slow speed of action. In both cases, this prevented us from detecting a significant effect on transmitter release. However, analogs that have strong efficacy but very slow kinetics of action (e.g. MF-06), are not likely to have significant effects at synapses during action potentials, but likely would increase transmitter release during non-physiological prolonged depolarizations (e.g. during prolonged voltage clamp steps or high potassium stimulation used in some experimental conditions; see section 4.6 below). In addition, such analogs may display stronger effects at synapses if combined with potassium channel blockers such as 3,4-DAP, which broaden the length of depolarization during an action potential. Of course, reduced cdk activity is also important if these compounds are to be advanced for the treatment of neuromuscular weakness. In this context, the highest potency and efficacy for any of our analogs was observed with MF-06; however, this advantage was offset by the retention of strong cdk inhibition (see Fig. 8). Overall, we strive to continue to refine the SAR of these compounds with the goal of developing new therapeutic leads.

4.6. Cav2 agonists as experimental tool compounds

Our newly developed Cav2 agonist gating modifiers create the opportunity to study calcium channels and calcium-dependent effects within cells in new and potentially interesting ways. Because the potential for use as tool compounds is not restricted to efficacy during natural action potential stimuli, even our analogs with slower kinetics will be useful. In this context, MF-06 shows great potential to significantly increase VGCC mean open time for experiments that use non-physiological depolarizations (long voltage clamp depolarization steps or high potassium stimulation). If prior use of (R)-roscovitine by experimentalists is any indication, these novel analogs with greater potency and efficacy should be valuable tool compounds. Such tool use has previously aided the study of Cav2 single channel conductance (Weber et al., 2010), Cav2 VGCC structure and function (Yarotskyy et al., 2012), Cav2 VGCC permeation (Buraei et al., 2014), the calcium ion control of transmitter release from synapses (Wen et al., 2013; Satake and Imota, 2014), and calcium ion dynamics in non-neuronal cells (Tamma et al., 2013).

HIGHLIGHTS.

• GV-58 and novel analogs are first-in-class Cav2 voltage-gated calcium channel gating modifiers with therapeutic potential to treat neuromuscular weakness.

• These analogs have contributed to our understanding of the structure-activity relationship of the GV-58 scaffold for action at voltage-gated calcium channels.

• Cav2 gating modifiers that can act quickly during brief action potentials and have strong potency and efficacy have the most promising therapeutic potential.

• These novel agonist analogs will also be useful tools for the study of calcium channel gating and Cav2 calcium channel-dependent pathways in cells.