Molecular simulations study of novel 1,4-dihydropyridines derivatives with a high selectivity for Cav3.1 calcium channel

Abstract

1,4-Dihydropyridines (DHPs) have been developed to treat hypertension, angina, and nerve system disease. They are thought to mainly target the L-type calcium channels, but low selectivity prompts them to block Cav1.2 and Cav3.1 channels simultaneously. Recently, some novel DHPs with different hydrophobic groups have been synthesized and among them M12 has a higher selectivity for Cav3.1. However, the structural information about Cav3.1-DHPs complexes is not available in the experiment. Thus, we combined homology modeling, molecular docking, molecular dynamics simulations, and binding free energy calculations to quantitatively elucidate the inhibition mechanism of DHPs. The calculated results indicate that our model is in excellent agreement with experimental results. On the basis of conformational analysis, we identify the main interactions between DHPs and calcium channels and further elaborate on the different selectivity of ligands from the micro perspective. In conjunction with energy distribution, we propose that the binding sites of Cav3.1-DHPs is characterized by several interspersed hydrophobic amino acid residues on the IIIS6 and IVS6 segments. We also speculate the favorable function groups on prospective DHPs. Besides, our model provides important information for further mutagenesis experiments.

Introduction

Voltage-gated calcium channels modulate the influx of calcium ions as efficiently as possible and regulate calcium ion concentration in the cell.^1–3 Defective calcium channels are implicated in many human diseases such as hypokalemic periodic paralysis, malignant hyperthermia, and epilepsy.^2,4–6

Calcium channels can be divided into high-voltage-activated (HVA) and low-voltage-activated (LVA) calcium channels.⁷ All the L-type calcium channels (LTCC) including Cav1.2 belong to HVA and all the T-type calcium channels (TTCC) including Cav3.1 belong to LVA. Cav1.2 channel is critical to brain and central nervous system.^8,9 It has been widely and deeply studied on architectural characteristics, biochemical natures, and physiological features.^10–13 Some DHPs derivatives, such as nimodipine¹⁴ and amlodipine,¹⁵ have been approved by Food and Drug Administration (FDA). Many drugs have already been worldwide used to treat cardiovascular diseases clinically by blocking the Cav1.2 channels.^16,17 Cav3.1 channel is widely spread among various tissues of mammals and dominate neuronal excitability, so it is vital to adrenal, cardiac, and systemic–toxic function.^6,18–20 Many notable Cav3.1 blockers have been detailed studied in experiment.²¹ Although fundamental research on the Cav3.1 channel has made great progress, its pharmacology is not well characterized²² and reports about structural model are very few.

DHPs are thought to primary target the LTCC,²³ but most of them block both the Cav1.2 and Cav3.1 channels concurrently. To solve the low selectivity of traditional calcium antagonist, many novel DHPs derivatives have been synthesized (Table1). Since we choose the (s)-p-nitrophenol plane as reference, the right side of the DHPs ring is referred to as the port side.²⁴ Bladen has studied the different affinity between these DHPs molecules and two channels by experimental means. Among of them, M12 exhibits a strong selectivity for Cav3.1 and acts quickly with an almost complete recovery, so it satisfies the ideal properties of drug molecules.²⁵ Consequently, we decide to carry out molecular simulations to quantitatively elucidate the above phenomenon.

Table 1.

Twelve Compounds and Four Available Values

Series	Structureb	Ligand	R			IC50 (μM)a
1		M1
		M2
		M3
		M4				4(Cav1.2)/2.4(Cav3.1)
		M5
2		M6
		M7
		M8
		M9
		M10
3
		M11
		M12				>30(Cav1.2)/3.2(Cav3.1)

^a values show in the far right column.

^bWe apply red numbers to annotate the carbonyl group on the structural diagrams for convenient expression.

In addition, the experiment revealed that the binding sites between Cav1.2 and DHPs located at IIIS5, IIIP, IIIS6, IVP, and IVS6 segments of thesubunits.^26–31 Conversely, up to now, binding sites between Cav3.1 and DHPs remains unclear. Therefore, determining binding sites of Cav3.1–DHPs complexes has clinical benefits for the treatment of human diseases, especially the nervous system disorders.

In view of the unavailable Cav3.1 crystal structure and the low sequence identity with known potassium or sodium channels, we built our Cav3.1 structure based on the Stary's open Cav1.2 model.¹¹ Molecular docking was performed to search initial ligand conformation. MD simulations were executed in an authentic biological environment which contains protein, ligand, lipids, water, and ions. Subsequently, we used linear interaction energy (LIE) method^32,33 to calculate the absolute binding free energy. Notably, the calculated results were consistent with the available experimental data. In the conformational analysis of ligands, we explored 12 derivatives, focusing on studying the substituent (hydroxyl group or benzoyloxy group) at the para-position of nitro, to investigate why first series (M1–M5) generally has a higher binding free energy than second series (M6–M10). Meanwhile, we also study the different R groups at port side of ligand and confirm the different orientations. The structure–activity relationship of ligands was deduced. The hydrophobic effect of M12 in two channels is detailed analyzed in this article. Binding sites between Cav3.1 and DHPs were preliminarily determined by MD simulations. In order to further identify these binding sites from energy distribution, we decomposed the total binding free energy into contributions per residue. Throughout the discussion section, we also speculated some favorable functional groups which can be used to design the new pharmacophore for LTCC and TTCC.

Results

Structural stability

The alignment, modeling, validation, and MD simulation of Cav1.2 model have been discussed in detail.¹¹ Like the Cav1.2 model, the initial clockwise Cav3.1 structure also arranges its four domains symmetrically around the central pore and exhibits pretty stable during MD simulation [Fig. 1(A)]. Four domains show the overall stability of the scaffold. Remarkably, the backbone of IS6, IIS6, IIIS6, and IVS6 segments also maintains fairly stable with RMSD values of 0.21, 0.21, 0.25, and 0.25 nm, respectively [Fig. 1(B)]. It is worth noticing that the selectivity filter (EEDD) of Cav3.1 has two characteristic Asp amino acids with a shorter side chain. They constitute an asymmetric trapezoid with two Glu amino acids in Cav3.1, while four same Glu amino acids form the square selectivity filter of Cav1.2 (Supporting Information, Fig. S2A). P loops also remain relatively stable during MD simulation. Especially, IP, IIP, and IIIP loops have a minimal fluctuation from 0.1 to 0.15 nm (Supporting Information, Fig. S2B). In short, our whole Cav3.1 model is reliable for further molecular docking.

Figure 1.

(A) Side view of Cav3.1 model. (B) Stability of S6 segments during MD simulations. Rmsd of four S6 segments. Superpositions of initial structure (gray) and average structure (rainbow).

RMSD values reach equilibrium after 20 ns with a range from 0.30 to 0.45 nm in Cav1.2–DHPs complexes and from 0.35 to 0.49 nm in Cav3.1–DHPs complexes (Supporting Information, Fig. S3). High structural stability is also corroborated by the small displacement of residues during MD simulation. The whole conformation of average structure keeps good consistency with the initial docked structure for both Cav1.2–DHPs and Cav3.1–DHPs complexes (Fig. 2). Ligands are coordinated by P loops, IIIS6, and IVS6 segments and undergo a slight translational motion due to the adjustment of protein. Actually, the small magnitude of spontaneous adjustments as well as the identification of the segments that show slight fluctuations is presumably caused by the shift of potential function from stochastic global optimization in docking to GROMOS96-53a6 in MD. The alpha-C of certain residues on IIIS6 and IVS6 segments almost hold steady when combined with ligands, further validating the stability of the inner pore region. Overall, these results substantiate our reasonable molecular docking and the stable complex structures.

Figure 2.

Side view of Cav1.2-M12 (A) and Cav3.1-M12 (B). Superpositions of initial docked structure (gray with green M12) and average structure (rainbow with salmon M12) of MD simulation.

LIE calculations

Experiments reveal four available IC50 values, including Cav1.2-M4, Cav1.2-M12, Cav3.1-M4, and Cav3.1-M12 (Table1), and provide a current blockage map for all compounds. We calculated the absolute binding free energy with the LIE method by extracting Lennard–Jones and Coulomb potential energy terms from stable stage of MD trajectory, in the sense that fluctuant and unreasonable sections were discarded. Based on aforesaid and, we used an additional constant term of −4.77 for Cav1.2–DHPs and −5.01 for Cav3.1–DHPs to obtain absolute calculated binding free energy. We applied the same for same channel, so LIE model has no free parameters affecting binding free energy.³⁴ Different reflects the different hydrophobic property of two channels, which will be discussed below.

Table2 depicts calculated binding energies of 24 complexes and four available experimental binding energies. The calculation with LIE method for M4 and M12 gave −7.16, −6.60, −7.93, and −7.23, deviated from the experimental value of −7.41, −6.21, −7.71, and −7.54 by 0.25, 0.40, 0.22, and 0.31, in excellent agreement with experimental results.²⁵ Considering the fact that error of these four complexes is lower than 0.40, our LIE model can estimate the binding free energy accurately. Intriguingly, the calculated results show that a ligand with benzyl or isobutyl at the port side (M3 and M4) has a higher binding energy than others. The particular M12 has a strong affinity with Cav3.1 model. Besides, our LIE model discriminates the moderate high-affinity group of M4 from the low-affinity group of M6 in Cav1.2–DHPs complexes, as well as in Cav3.1–DHPs complexes. Although there is a slight overestimation or underestimation for several complexes (Cav1.2-M11, Cav3.1-M7, and Cav3.1-M11), the LIE method can still provide correct binding free energies for most complexes. In a nutshell, LIE method which has been applied to many complexes successfully^34–36 also can evaluate and rank the binding free energy of our complexes excellently.

Table 2.

Average Calculated LIE Binding Free Energies () for the 24 Complexes^a

Channel	Ligand							Error	ΔGLIE (%)
Cav1.2	M1	−30.63 ± 0.21	−48.65 ± 0.73	−15.91 ± 0.18	−11.65 ± 0.34	−6.46 ± 0.67			73.84%
	M2	−31.84 ± 0.10	−53.70 ± 0.26	−16.21 ± 0.39	−11.84 ± 0.47	−7.11 ± 0.31			67.09%
	M3	−37.65 ± 0.25	−59.78 ± 0.45	−15.14 ± 0.14	−10.39 ± 0.66	−7.02 ± 0.56			67.95%
	M4	−34.03 ± 0.18	−55.12 ± 0.32	−16.01 ± 0.29	−12.14 ± 0.54	−7.16 ± 0.58	−7.41	−0.25	66.62%
	M5	−33.52 ± 0.26	−49.42 ± 0.50	−15.04 ± 0.04	−9.47 ± 0.70	−5.59 ± 0.49			85.33%
	M6	−40.01 ± 0.05	−59.47 ± 0.13	−12.23 ± 0.17	−6.32 ± 0.35	−5.75 ± 0.24			82.96%
	M7	−40.95 ± 0.11	−60.08 ± 0.20	−12.32 ± 0.07	−7.25 ± 0.94	−6.05 ± 0.45			78.84%
	M8	−46.06 ± 0.23	−64.49 ± 0.27	−11.63 ± 0.15	−8.06 ± 1.18	−6.57 ± 0.53			72.60%
	M9	−42.30 ± 0.04	−65.78 ± 0.47	−9.74 ± 0.28	−4.26 ± 0.42	−6.66 ± 0.38			71.62%
	M10	−42.23 ± 0.09	−68.13 ± 0.59	−12.76 ± 0.26	−5.54 ± 0.85	−6.35 ± 0.64			75.12%
	M11	−33.93 ± 0.14	−51.04 ± 1.24	−11.18 ± 0.10	−9.67 ± 1.49	−7.22 ± 0.92			66.07%
	M12	−37.40 ± 0.03	−55.73 ± 0.50	−10.62 ± 0.07	−7.17 ± 0.62	−6.61 ± 0.74	−6.21	0. 40	72.27%
Cav3.1	M1	−30.87 ± 0.27	−44.90 ± 0.41	−14.81 ± 0.22	−11.78 ± 0.63	−6.43 ± 0.57			77.92%
	M2	−32.00 ± 0.05	−47.70 ± 1.33	−15.91 ± 0.06	−13.20 ± 0.72	−6.85 ± 0.40			73.14%
	M3	−37.60 ± 0.12	−61.70 ± 0.26	−15.44 ± 0.18	−9.84 ± 0.69	−7.30 ± 0.32			68.63%
	M4	−33.97 ± 0.16	−54.57 ± 0.39	−15.76 ± 0.31	−13.58 ± 0.51	−7.93 ± 0.46	−7.71	0.22	63.18%
	M5	−33.91 ± 0.02	−51.42 ± 0.38	−15.89 ± 0.19	−11.05 ± 0.82	−6.39 ± 0.53			78.40%
	M6	−40.09 ± 0.17	−60.75 ± 0.43	−12.51 ± 0.20	−5.10 ± 0.37	−5.56 ± 0.25			90.10%
	M7	−40.89 ± 0.04	−61.07 ± 0.68	−11.96 ± 0.36	−8.23 ± 1.01	−7.06 ± 1.05			70.96%
	M8	−45.54 ± 0.13	−69.34 ± 0.36	−10.80 ± 0.03	−4.48 ± 0.56	−6.60 ± 0.37			75.91%
	M9	−42.08 ± 0.21	−61.37 ± 1.06	−11.22 ± 0.11	−6.38 ± 0.44	−6.42 ± 0.68			78.04%
	M10	−42.41 ± 0.08	−60.93 ± 0.54	−11.78 ± 0.05	−7.51 ± 1.28	−6.53 ± 0.72			76.72%
	M11	−33.77 ± 0.15	−50.30 ± 0.59	−10.77 ± 0.23	−4.09 ± 0.61	−5.13 ± 0.43			97.66%
	M12	−38.05 ± 0.06	−57.21 ± 0.77	−9.24 ± 0.20	−6.34 ±0.75	−7.23 ± 0.36	−7.54	−0.31	69.29%

^aAverage ligand-surrounding Vander Waals () and electrostatic () energies for the bound (b) and free (f) states.

^bThe calculated binding free energies are calculated by LIE method using Eq. (2).

^cThe experimental binding free energies are obtained using Eq. (1) based on the experimental data. The percent of in calculated binding free energies is shown in the far right column.

Locating the DHPs in the crevice between IIIS6 and IVS6 segments

All the ligands are postulated to slip into a semisurrounded annular crevice constituted by IIIS6 and IVS6 segments. Ligands in Cav1.2 are restricted by the side chain of Y644 and 861Y on the top, by I648 and A865 on the middle and by the hydrophobic side chain of M653 and I869 on the bottom [Fig. 3(A) and (C)]. It shares some common features with model of Lipkind and Fozzard.³⁷ In a similar way, ligands in Cav3.1 are restricted by the hydrophobic side chain of F272 and F370 on the top, by V276 and A374 on the middle and by the hydrophobic side chain of L281 and L378 on the bottom [Figs. 3(B,D) and Fig. 4]. Apparently, homology Cav3.1 model shares completely analogous features with Cav1.2 model. The statistics on distance also confirm the location.

Figure 3.

Inside view of Cav1.2-M2 (A), Cav3.1-M4 (B), Cav1.2-M6 (C), and Cav3.1-M6 (D). Ligand (salmon) locates in the pocket constituted by IIIS6 segment (magentas) and IVS6 segment (yellow). Polar interactions (black-dashed line), part of IIIP loops (red) and IVP loops (sand) show in the picture.

Figure 4.

Orientation of M3 (A) and M8 (B) in the pocket of Cav3.1. The dashed line represents the residue behind.

Interaction between DHPs and calcium channels

According to the different substituent on the para-position of nitro, ligands (M1–M10) are divided into first and second series. The particular M12 in third series has a high selectivity for Cav3.1 channel. All ligands locate in the inner crevice of protein and involve interactions with the P loop, IIIS6 and IVS6 segments. The interactions dictate the torsion of the important residues and confirm the unambiguous orientation of the ligand, and then determine the binding sites preliminarily. We will make an exhaustive analysis of principal interactions to elucidate the precise binding modes between these ligands and calcium channels.

First series with a hydroxyl group

In Cav1.2–DHPs, polar residue Y644 twists about 30° to impose restriction on the ligand by generating a clamp with N870. The adjacent residue I645 establishes not only a polar contact but also powerful hydrophobic effect with ligand. Isobutyl of I869 deflects toward the ligand while I648 and M653 almost remain motionless during MD simulation. Without exception, these nonpolar residues provide sufficient hydrophobic effects [Fig. 3(A)]. From perspective of ligand, the suitable-sized isobutyl group of M4 allows the molecule to maintain an appropriate distance with binding sites and emerges as a higher binding free energy. Dramatically, 2-carboyl moiety of M4 also plays a fundamental role in structural stability because it develops polar interactions and H-bond with residues. On the contrary, thumbnail methyl of M1 and oversized tert-butyl of M5 keep themselves away from binding sites, which give rise to the devoid of above interactions and produce their low binding free energy.

In Cav3.1–DHPs, nonpolar residue F272 twists about 28° toward the periplasm side to coordinate the ligand with L378. Benzene ring of hydrophobic residue F370 is almost vertical to the pore central axis and forms a pi–pi conjugation effect with p-nitrophenol plane. Isopropyl group of V377 deflects inward the pore region to interact with isobutyl of ligand [Fig. 3(B)]. Moreover, first series obtains dominant orientation in the Cav3.1 with hexahydroquinoline ring being nearly perpendicular to selectivity filter and NH moiety being far away from the alkaline guanidine group of R340. Nevertheless, the displacement of corresponding residues (F618 in Cav1.2→K246 in Cav3.1) on P loops produces a disadvantageous polar effect, potentially accounting for the ubiquitous lower free energy in all three series of Cav3.1–DHPs complexes.

Second series with a benzoyloxy group

In contrast to the small hydroxyl substituent in first series, the big and inflexible benzoyloxy substituent in second series is associated with an expanded and squashed pocket. IIIS6 and IVS6 segments splay outward concomitantly to accommodate the ligand [Fig. 3(C,D)].

In Cav1.2–DHPs, the “wall corner” constituted by residues Y861, M862, and A865 from three dimensions powerfully potentiate the ligand-binding capacity. Hydrophobic residue F618 provides its benzene ring for p-nitrophenol plane as a shield from the influence of polar solvent. Whereas, as the bottom of pocket, polar N870 kinks completely and its positive-charged area repel two methyl groups of ligand. Y644 and I645 have a further distance with ligands and miss partial interactions. Above adjustment triggers the different orientation of second series as well as the absence of important interactions.

In Cav3.1–DHPs, the side chain of F272 bends downward and forms a clamp with L378 to restrain the ligand. The adjustment promotes 1-carboxyl moiety to establish a somewhat feeble but persistent polar interaction with V280. Interactions involved in nitro group are not perceived. Besides, oversized volume of M8 dictates the orientation, where its hexahydroquinoline ring protrudes into the central pore with an additional inclination about 25° than M6 [Fig. 4(B)]. Positively charged region of guanidine group in R340 is mutually exclusive with NH moiety of ligands. Thus, R340 has a negligible effect on first series but a detrimental effect on second series. These reasons lead to the lower binding free energies of second series, rendering these ligands less potent.

As a result, the interaction with the protein is more favorable for a hydroxyl group (first series) than for a benzoyloxy group (second series), which is consistent with experimental data. Meanwhile, conformational analysis exhibits that ligands occupy two alternative orientations in the pocket. For the first series with an optimal orientation, nitro group adjoins with IIIS6 segment and port side faces toward the cytoplasmic side. On the contrary, for the second series, nitro group is adjacent to IVS6 segment and port side faces toward the periplasmic side (Figs. 3 and 4).

M12 with a high selectivity for Cav3.1

In comparison with others, M12 has a unique structural feature with a higher selectivity for Cav3.1 and acts quickly with an almost complete recovery, so it has potential to be a new generation of selective calcium antagonist. Hence, we make a thorough inquiry of the hydrophobic effect between M12 and channels.

M12 locates in a tabular pocket of Cav1.2 with an unfavorable orientation (Fig. 5). Two methyl groups are implicated in the hydrophobic interactions with nonpolar residues I645 and I648, and polar effect with hydrophilic residue Y644. Trifluoromethyl group interacts with the Y861 and M862 of the IVS6 segment. However, hexahydroquinoline ring is unable to penetrate into the bottom of the pocket. Mutual exclusion is detected between positive charged region of pyridine ring and amidogen of hydrophilic N870 (“1” in Fig. 5). Meanwhile, NH moiety of ligand also suffers adverse impact from the hydrophilic N654 (“2” in Fig. 5). 2-Carbonyl moiety and pyridine ring are partial accessible to solvent without the sequestration of hydrophobic residue. Therefore, binding energy decreases significantly.

Figure 5.

Hydrophobic surfaces of Cav1.2-M12 and Cav3.1-M12 complexes. To observe the pocket more clarity, we only retained the IIIS6 and IVS6 segments. Along with the change of the colors from dodger blue to orange red, hydrophobic character gradually enhanced. Chimera³⁸ has been used to generate the figure.

The overall M12 locates in a deeper pocket of Cav3.1 with a favorable orientation (Fig. 5). Two methyl groups are surrounded by V280, L281 on the left, by F370 in the rear and by A374 on the right. These ambient nonpolar residues potentiate the great hydrophobic effects with ligand. Trifluoromethyl group presses close to the F272 and L273 of IIIS6 segment. Obviously, hexahydroquinoline ring was whole accommodated by the larger crevice between V276, A277 of IIIS6 segment and A374 of IVS6 segment. Pyridine ring stretches toward high hydrophobic residues V377, L378, and V381 (“i” in Fig. 5) and is comprehensively protected by these residues from the adverse effect of polar solvent. Charge distribution reveals a mutual attraction between its positive-charged region and negative-charged area of L378. Ultimately, M12 has a stronger binding affinity with Cav3.1 than Cav1.2. Our molecular simulation verifies the hypothesis proposed by Bladen et al.²⁵

Dramatically, 18 hydrophobic residues locate at the pocket region in Cav3.1 channel with a more negative (−5.01) while the value is 17 in Cav1.2 channel with a slight negative (−4.77). The shift of binding sites from hydrophobic residue Phe in Cav3.1 to hydrophilic residue Tyr in Cav1.2 offers a direct explanation for the different (Supporting Information, Fig. S1 and Table2). The correlation relationship coincides with previous research quite remarkable.³⁹ The paramount constant term occupies a large percentage (>63%) of binding free energy. Hence, it implies that hydrophobic effects play a predominant role in both Cav1.2–DHPs and Cav3.1–DHPs complexes.

Binding sites between Cav3.1 and DHPs

We measured the RMSF of Cav3.1 and Cav3.1–DHPs complexes. The data reveals that the alpha-C of aforementioned residues has very little fluctuations when combined with ligands (Supporting Information, Fig. S4A). Although there is partial and temporary torsion on some residues, the scaffold of complexes remains stable during MD simulation. The similar phenomenon also exists in Cav1.2–DHPs (Supporting Information, Fig. S4B). Therefore, we decided to further identify the binding sites in conjunction with energy distribution.

We decomposed the total binding free energy into contributions per residue. Then we analyzed the individual residue energetic distribution comprehensively. Binding sites of Cav3.1-DHPs complexes are characterized by interspersed hydrophobic amino acid residues, containing F272, L273, V276, A277, V280, L281 of IIIS6 segment and F370, A374, V377, L378 of IVS6 segment (Fig. 6). They participate in forming the pocket or region wall to accommodate inhibitors and make tremendous contributions to binding free energy. Moreover, hydrophilic K246 and R340 are condemned to reduce the binding free energy.

Figure 6.

Average residue–ligand binding free energies distribution of Cav3.1-DHPs complexes. Magenta (IIIS6) and yellow (IVS6) bars represent the most favorable contribution to binding free energy while red (K246) and sand (R340) bars unfavorable.

Meanwhile, we also execute energy decomposition for Cav1.2–DHPs complexes. Y644, I645, I648, and M653 of IIIS6 segment, Y861, M862, A865, I869, and N870 of IVS6 segment play a pivotal role in the binding free energy (Supporting Information, Fig. S5). These amino acid residues basically agree well with experimental binding sites.^26,28

Discussion

This work is based on a report concerning the synthesis and evaluation of a new series of calcium channel inhibitors. In this investigation, we built a homology model of Cav3.1 for molecular docking and executed MD simulation, followed by absolute binding free energy evaluation and decomposition.

Even though partial and temporary torsion occur on some residues, the scaffold of complexes hold steady during MD simulation. We calculated the binding free energies of 24 complexes by extracting the stable MD trajectories. Calculated absolute binding free energies revealed an excellent consistency with the available experimental values with a mean unassigned error smaller than 0.40. We validated the major interactions and orientation of ligands in the conformational analysis phase. Then energy distributions further identified binding sites of Cav3.1–DHPs.

The hydrophobic effects produced by nonpolar residues play a decisive role in binding free energies of all complexes. Many polar interactions involved in nitro group and carbonyl moiety also distribute in some complexes. We concluded that the R group on the port side results in the different binding energy within the same series. The substituent on the para-position of nitro group results in the different binding energy between different series (Tables1 and II). In addition, ligands occupy two alternative orientations in the pocket.

Based on the results of this article, we can make a preliminary forecast for prospective DHPs drug. The ideal DHPs molecule prefers a hydroxyl substituent at the para-position of nitro. The momentous R groups at the port side can be the benzyl and isobutyl. The molecule will obtain an optimum orientation in calcium channels and exhibit stronger potency than traditional drugs. Conspicuously, M12 has a higher selectivity for Cav3.1 channel and is expected to be a prominent selective calcium antagonist. We expect that our investigation could help researchers to design new pharmacophore of calcium channel blockers.

Binding sites of Cav3.1–DHPs are deemed to be F272, L273, V276, A277, V280, L281 of IIIS6 segment and F370, A374, V377, L378 of IVS6 segment. They are in a location similar to the binding sites of Cav1.2–DHPs (Supporting Information, Fig. S1).²⁶ Developing new selective drug is extremely urgent and it is imperative to apply an exhaustive Cav3.1 alanine scanning study of DHPs blockade. It would be meaningful to identify binding sites in the experiment and estimate our theoretical results. DHPs produce a reduced affinity when Y644, I648, M653, and N870 of Cav1.2 channel is mutated.^29,30 We predict that the mutation of F272, V276, L281, F370, L378 of Cav3.1 channel will seriously weaken the binding affinity. Moreover, another noteworthy element is that A277 and A374 are perceived to control the orientation of adjacent F272 and F370. Since these two Phe constitute the pocket and interact with ligand (Fig. 6), so mutation of these two Ala may alter original shape of pocket and change the binding free energy.

Methods

Homology modeling

Models used in this study are pore region of Cav1.2 channel (accession number: P15381) and Cav3.1 channel (accession number: O43497). Thus far, crystal structure of Cav1.2 is not available. We used the open Cav1.2 model built by Stary et al.¹¹ In general, the higher sequence identity between template and target, the more accurate are the model. Cav3.1 appears less homology (<28%) with known three-dimensional potassium or sodium channels. By contrast, its pore region (S5-P-S6) shares high-sequence identity with Cav1.2 (>40%). Therefore, we selected clockwise Cav1.2 model as a template to build Cav3.1 structure by homology modeling. This approach has been applied successfully.⁴⁰ We applied CLUSTAL X 2.1⁴¹ to calculate sequence alignment. Several highly conserved amino acids ascertain the alignment of these segments (Supporting Information, Fig. S1). In particular, F and W of P loop permit the widely spread Glu to align with Asp in both IIIP and IVP loops of Cav3.1. Modeller 9.13⁴² was used to generate and optimize a 3D Cav3.1 model. The MD simulation was performed with Gromacs-4.5.3⁴³ using the GROMOS96-53a6 force field.⁴⁴ Model was embedded in a cubic box which is stuffed with POPC lipid bilayer^45,46 and equilibrated system was solvated with SPC water. Further, Na and Cl ions were added randomly to neutralize the system and generate a 100 mM ionic strength. After energy minimization and positional restrained MD, a 100 ns unrestrained MD simulation was executed to verify the stability of our equilibrium Cav3.1 model.

Docking calculations

According to the different substituent at the para-position of nitro, the studied 12 DHPs derivatives can be divided into three series (Table1): compounds with a hydroxyl group (M1–M5), compounds with a benzoyloxy group (M6–M10), and other compounds with a similar scaffold (M11–M12). In first and second series, different hydrophobic R groups at port side are used to distinguish each other within same series. All the compounds are electrically neutral. The geometric parameters and 3D configurations of the ligands were produced by Gaussian-09 package⁴⁷ with B3LYP method using 6-31g basis sets. Molecules obtained the stable conformation via energy minimization. Subsequently, Autodock Vina⁴⁸ using a stochastic global optimization of the scoring function was performed to dock the ligands into the inner pore region of calcium channels. We implemented semiflexible docking for both Cav1.2 and Cav3.1. Structural stability during MD simulation, as well as a small RMSD value, is considered as a criterion to select preferred conformation. Thus 24 ligand–protein combining structures were screened out from 216 alternative conformations.

MD simulations

The topology files of ligands were generated on the PRODRG server.⁴⁹ The topology files of proteins were generated using GROMACS package with the GROMOS96-53A6 force field. MD simulation was performed with Gromacs-4.5.3. All the simulations repeated three times.

For the ligand–protein simulation (bound states), complexes were embedded in an cubic box of POPC lipid bilayer using the INFLATEGRO tools.⁵⁰ After the removal of overlapping lipids and 22 iterations of compression with energy minimization, the system reached equilibrium with an area per lipid of 74 Ų. Then it was solvated with SPC explicit solvent, and we obtained a 100 mM ionic strength of NaCl by adding sufficient counterions. The energy was minimized using the steepest descent algorithm, followed by positional restrained MD which includes a 100 ps isochoric–isothermal (NVT) simulation to 300 K and a 1 ns isothermal–isobaric (NPT) simulation to 1 atm. All heavy atoms were restrained with the force constant of 1000. Finally, we released the position restraints and ran a 100 ns production MD for data collection with a time step of 2 fs. Protein–ligand potential energies were saved every 4 fs. LINCs algorithm⁵¹ was applied to constrain all bonds. The temperature and pressure were maintained using Nose–Hoover thermostat⁵² and Parrinello–Rahman pressostat method⁵³ with a time constant of 0.3 and 1.0 ps, respectively. The short-range interactions were calculated with the cutoff of 1 nm, while the long-range electrostatic interactions were computed using Particle Mesh Ewald summation method.⁵⁴

Meanwhile, for the ligand in solvent simulation (free states), ligand was placed in a cubic box of SPC water. The system was energy minimized using steepest descent algorithm. A 5 ns unrestrained and production MD simulation was carried out with energies collection every 4 fs. The temperature and pressure arrived at 300 K and 1 atm using the Nose–Hover thermostat and Parrinello–Rahman pressure coupling method.

Linear interaction estimation of binding free energies

We estimated the experimental binding energy using the common approximation equation:

101

where and are the gas constant and the experimental temperature. Based on the simulations of thermodynamic relevant states and the linear response assumption, the computational absolute binding free energy was calculated with the widely used LIE method.³² In general, it uses simulations of bound states and free states to calculate via following Eq. (2):

1

whereand refer to average energies of ligand-surrounding Vander Waals (Lennard–Jones) and electrostatic parts during MD simulation. denotes the difference of average energy between the bound and free states. Different weight coefficients and represent the nonpolar and polar contribution to the binding energy, respectively. Parameter was empirically set to 0.181.⁵⁵ Parameter which was derived from the linear response approximation depends on the chemical nature of the ligand. First series was classified as dipolar compounds with a hydroxyl group ( = 0.37); Second and third series were categorized as dipolar compounds without hydroxyl group ( = 0.43).^32,55 In LIE models, the constant term is used for modifying molecular mechanics energy to obtain the absolute binding free energies. It has been proved to be intimately related to hydrophobic nature of the binding sites.³⁹

Glossary

DHPs

1,4-Dihydropyridines

HVA

high-voltage-activated

LIE

linear interaction energy

LTCC

L-type Calcium Channels

LVA

low-voltage-activated

MD

molecular dynamics

TTCC

T-type Calcium Channels

Supporting Information

Additional Supporting Information may be found in the online version of this article.

Supporting Information