Synthesis, Cytotoxicity Evaluation, and Computational Insights of Novel 1,4-Diazepane-Based Sigma Ligands

Abstract

Among several potential applications, sigma receptor ligands can be used as antipsychotics, antiamnesics, and against other neurodegenerative disorders as well as neuroprotective agents. We present herein a new series of diazepane-containing derivatives as σR ligands obtained by a conformational expansion approach of our previously synthesized piperidine-based compounds. The best results were reached by benzofurane 2c, 3c and quinoline 2d, 3d-substituted diazepane derivatives, which showed the highest σR affinity. The cytotoxic activities of synthesized compounds were evaluated against two cancer cell lines, and the results indicated that none of the compounds induced significant toxicity in these cells. We also evaluated the antioxidant activity by radical scavenging capacity of our best compounds on ABTS and H₂O₂. The results obtained reveal that our new derivatives possess an excellent antioxidant profile and could be protective for the cells. Overall, the benzofurane derivative 2c due to its strong interaction with the active site of the receptor, as confirmed by molecular dynamic simulations, emerged as the optimum compound with high σ1R affinity, low cytotoxicity, and a potent antioxidant activity.

The sigma receptors (σR) are a class of proteins initially classified, by Martin and co-workers,¹ as a subtype of the opiate receptors. Further studies revealed them to be a different receptor class comprising two distinct subtypes: σ1 and σ2.²⁻⁵ The σ1R is a chaperone protein, cloned in 1996 from several tissues including human, consisting of 223 amino acids^6,7 with a MW of 25.3 kDa.⁸ Crystallized 20 years later, it revealed a trimeric protein organization.⁹ The σ1R subtype is primarily localized to mitochondria-associated ER membranes (MAM) of neuronal and peripheral cells, such as cardiac myocytes and hepatocytes. This receptor can also translocate to the plasma membrane or ER-membrane and regulate the activity of other proteins by modulating different ionic channels via an IP₃-indipendent mechanism.^10,11 The σ1Rs have neuroprotective and antiamnesic activity^12,13 and modulate opioid analgesia¹⁴ as well as drug addiction,¹⁵ and their antagonists seem to be effective against the negative manifestations of schizophrenia without producing extrapyramidal side effects.^16,17 In addition, several studies suggest a role for σ1R in tumor biology, since its expression increased in some cancers.¹⁸

After 40 years from the discovery of σRs,¹ in 2017, the σ2R subtype has been purified and identified as transmembrane protein-97 (TMEM97),¹⁹ an endoplasmic-reticulum-resident transmembrane molecule implicated in cholesterol homeostasis due to its association with the lysosomal transporter NPC1.^20,21 The σ2R crystal structure is still elusive, but several pharmacophore models have been proposed.²²⁻²⁵ The σ2Rs are overexpressed in many cancer cell lines including lung cancer,^26,27 breast cancer,²⁸ ovarian cancer,²⁹ glioma cancer,³⁰ and gastric cancer.³¹ In this context, since σ2R agonists can induce tumor cell death, they have been proposed as potential antitumor drugs. On the other hand, the σ2Rs are widely expressed in cerebellum, red nucleus, and substantia nigra and are a potential target for the treatment of movement disorders and of neuroleptic-induced acute dystonia.³² In addition, σ1R antagonists as well as σ2R agonists can modulate neuropathic pain.^33,34

In the past decade, our group has synthesized and biologically evaluated an extensive series of compounds both of preferential affinity for σ1R and σ2R subtypes. Following up our studies in this field, we report herein the development of a new class of sigma ligands designed through the expansion of the conformational selection paradigm applied to our previously synthesized piperidine-based σ1R ligands 1 (Figure 1).³⁵

Figure 1.

Conformational expansion approach starting from previously synthesized sigma ligands 1.

In addition to the spacer replacement, we opted to expand the new series of compounds by using various aromatic fragments, including heterocycles, both monocycles and bicycles, linked to the amide carbonyl group. Moreover, in order to verify the influence on the sigma affinity of the substitution on the benzyl moiety, we decided to retain the unsubstituted phenyl ring, present in many σ1R ligands and the 2,4-dimethyl substituted phenyl ring, typical of several σ2R ligands.³⁶ The synthesis of our new diazepane-based derivatives 2a–g and 3a–g is depicted in Scheme 1.

Scheme 1. Synthesis of Compounds 2a–g and 3a–g.

(a) DCM, Ar-COCl, Et₃N, 0 °C; (b) DCM, TFA, rt; (c) DCM, 2,4-dimethylbenzaldheyde, NaCNBH₃, rt.

The synthetic route of our new series of compounds (Supporting Information) was carried out by treating the appropriate, commercially available, 1-benzyl-1,4-diazepane, which was made to react with the appropriate aroyl chloride to give the corresponding first subseries 2a–g. These compounds did not need further purification after the classical workup. The 2,4-dimethyl derivatives 3a–g were obtained in three reaction steps, starting from 1-Boc-1,4-diazepane and the corresponding aroyl chloride to provide the acylated intermediates 4a–g. The cleavage of protecting N-Boc group with TFA led to the intermediates 5a–g which were subsequently N-alkylated, with a direct reductive amination using 2,4-dimetylbenzaldheyde and NaCNBH₃, to give the final subseries 3a–g. These compounds were purified by DCVC technique.

The σ1R and σ2R affinities of the test compounds were determined in competition experiments by radiometric assays, using [³H]-(+)-Pentazocine as radioligand for the σ1R assay and [³H]-DTG (di-o-tolylguanidine) as radioligand in the σ2R assay. Compounds 2a–g and 3a–g were tested against σ1R and σ2R of animal origin, prepared from guinea pig brain and rat liver membranes by homogenization, centrifugation, and washing of the respective tissues. We also performed a competition experiment toward GluN2b subunit containing NMDA receptors in a radioligand binding assay. This receptor subtype plays important roles in synaptic transmission and plasticity, learning, memory, and other physiological and pathological processes.^37,38 Hence, antagonists of the GluN2b subunit are of interest as neuroprotective drugs for various CNS disorders. The radioligand used in the competition assay was [³H]-labeled Ifenprodil, a prototypical allosteric inhibitor of the GluN2b subunit (Supporting Information).

For compounds with affinity value higher than 100 nM, only one measure was performed. The σ1R, σ2R, and GluN2b affinities of compounds 2a–g and 3a–g are presented in Table 1.

Table 1. Affinities of Compounds 2a–g and 3a–g toward σ1, σ2, and GluN2b Receptors.

^aOnly compounds with highest affinities (<100 nM) were tested in triplicate. For low-affinity compounds, the competition curves were recorded only once (single value), whereas the inhibition of the radioligand binding (shown as %) was assayed at a test compound concentration of 1 μM.

^bNot tested.

From the obtained data we can summarize the following: (i) the bulky diazepane spacer retained, or even improved, the σR affinity to both σ1 and σ2, with respect to the piperidine ring; (ii) only bicycle derivatives displayed moderate to high affinity toward both σR subtypes, while the corresponding monocycle analogues were weak inhibitors or avoiding of σR affinity; (iii) the best results against σ1R were reached by benzofurane derivative 2c, while its 2,4-dimethyl substituted analogue 3c gave the best pan-affinity with K_i values of 8.0 and 28 nM toward both σR subtypes and also the best GluN2b inhibition value of 59 nM; (iv) the 2,4-dimethyl substitution on benzyl moiety derivatives improves the σ2 over σ1 affinity of the bicycle derivatives, as well as the affinity toward the GluN2b subunit receptor.

To get insight into the interaction of our compounds into the σ1R active site, we performed a computational assessment of the best σ1R ligand of the series, 2c, in comparison with its monocycle analogue 2a.

We prepared the σ1R following our previous procedure³⁶ (Supporting Information) and we docked compounds 2a and 2c to the target by following the same protocol.

The comparison between the optimum pose obtained for each compound (Figure 2) suggests that compound 2c slides further into the pocket than 2a pushing its benzene ring to interact with Trp164 and Phe133, closed at the bottom by Tyr206, and forming an H-bond with Thr181. Moreover, compound 2c’s optimum pose is predicted to be in touch through 17 hydrophobic interactions with target residues and a hydrogen bond with Thr181 (inset in Figure 2a). Also compound 2a interacts with the target with 17 hydrophobic interactions including Thr181 and shares 14 of those interactions with compound 2c (inset in Figure 2b).

Figure 2.

3D putative of (a) compound 2c and (b) compound 2a in the optimum AutoDock pose. Protein residues interacting with both compounds by van der Waals interactions are highlighted in red, Thr 181 in green, other interacting residues (Trp 164, Phe 133, Tyr 206 in panel (a) and Thr202, Val84, Ile124 in panel (b)) are white. In the insets: schematic diagrams of the interaction between the receptor and the respective compounds in the same optimum AutoDock pose. Protein residues interacting with the compounds by van der Waals interactions are highlighted in red, while hydrogen bonds are indicated by green dotted lines. The hydrogen bond distances are also indicated. Residues interacting with both compounds through van der Waals interactions are circled in red. Thr 181 (circled in green) forms a hydrophobic interaction with compounds 2a, while a hydrogen bond with compound 2c.

To confirm the docking result and understand the different behaviors of the two compounds, we ran 250 ns of molecular dynamics (MD) simulation of the complexes in water solvent. The ligand topologies were built with ATB.³⁹ The topologies were validated as the molecular mechanics minimized structure of compound 2a had root-mean-square deviation (RMSD) of 0.01007 nm with respect to the semiempirical minimized structure, while for compound 2c the same RMSD was 0.0082 nm.

The trajectories, scored with the Autodock Vina scoring function,⁴⁰ showed a constant binding score for both compounds (Figure 3a,b). For both systems the protein backbone RMSD diverged along the dynamics up to 0.6 nm (Figure 3c,d), as expected by simulating only a monomer of the extracellular domain. The ligands RMSD with respect to the fixed protein backbone, below 0.4 nm for compound 2c (Figure 3c), revealed a major conformational change for compound 2a (RMSD > 1.0 nm, Figure 3d). These variations were not reflected in the protein backbone radius of gyration constant at 1.6 nm for both systems (Figure 3e,f). Larger protein rearrangements peak at residues 190–200 for both systems, as observed in the protein backbone root-mean-square fluctuation (RMSF > 0.5 nm, Figure 3g,h). These are part of the helical structure delimitating the pocket but not directly interacting with the ligands (Figure 3i,j and insets in Figure 2). Along the whole simulated time ligands were not observed to leave the pocket with compound 2c maintaining its position but rotating its benzene ring (Figure 3i) and compound 2a changing position by both flipping its orientation and further sliding inside the pocket (Figure 3j), a movement not associated with a gain in binding energy as calculated by Autodock Vina score (Figure 3b).

Figure 3.

Molecular dynamics simulations analysis for compound 2c (left column) and 2a (right column): (a,b) Autodock Vina score, (c,d) protein backbone RMSD (green) and ligands RMSD in the protein backbone frame (blue), (e,f) protein backbone radius of gyration, (g,h) protein backbone RMSF with values above 0.25 nm highlighted in magenta. All values measured with respect to the starting configuration correspond to the minimized optimum pose identified by docking. Simulation snapshots taken at the lowest Autodock Vina score (circled in panels a,b) for (i) compound 2c and (j) compound 2a. Protein residues with RMSF above 0.25 nm are highlighted (magenta). The starting ligand configuration is also indicated (white).

The MM/PBSA analysis confirmed that the van der Waals interactions were the major liable for binding the compounds, a contribution that is more relevant for compound 2c which also benefits of a less unfavorable polar solvation energy than that calculated for compound 2a (Figure 4a,b). This result is reflected by the single amino acid contribution to the binding energy. Indeed, for compound 2a several amino acids opposed to the binding with Arg119 and Glu172 contribution larger than 1 kcal/mol, followed by Gln135, Asp126, and His154. Instead, compound 2c is synergistically kept bound to its interacting site by several residues (Figure 4d). More in detail (Figure 4e,f) there are seven residues (namely Leu105, Ile124, Phe107, Trp89, Val84, Leu182, and Tyr103) contributing to the binding energy with more than 0.9 kcal/mol with the major contributing force to be ascribed to van der Waals forces and hydrophobic interactions.

Figure 4.

MMPBSA analysis: energetic contributions to the receptor binding of (a) compound 2c and (b) compound 2a, and amino acids contribution to the total binding energy of (c) compound 2a and (d) compound 2c with highlighted the amino acids opposed to the binding with predicted energy larger than 0.3 kcal/mol; (e) details of the contributions of each amino acid with binding energy larger than 0.9 kcal/mol for compound 2c; (f) snapshot of the complex σ1/2c with highlighted amino acids (shades of red) contributing to the binding energy with more than ±0.9 kcal/mol. All data averaged over the last 150 ns of the molecular dynamics trajectory.

Overall the larger ring keeps the ligand fixed in its position, which is more accessible to Thr181 for H-bonding. A larger number of contacts deep into the pocket further inhibited the molecular rearrangement inside the protein pocket.

The effects of this new set of σ1R ligands on cell health were evaluated by testing the cytotoxic response of the human pancreatic carcinoma (PANC1) and human neuroblastoma (SH-SY5Y) cell lines, selected because they express high levels of σ1R.¹⁸ To this aim, we selected the most interesting compounds (2c, 2d, 3c, and 3d) and tested their potential toxicity by MTT assay (Supporting Information, Table S1 and Table S2). The experiments revealed that none of our diazepane-containing derivatives showed significant cytotoxicity at different concentrations, with the exception of compound 3d, which exhibited a moderate toxicity toward PANC1 cells, but only at 100 μM concentration (viability of 51%). Interestingly, compounds 2c and 2d, which exhibited the best σ1R affinities (K_i = 8 and 19 nM, respectively), resulted to be the less toxic (viability: 127 and 196% at 50 μM; 98 and 127% at 100 μM, in SH-SY5Y and PANC1, respectively). Therefore, considering the general consensus that σ1R agonists promote cell survival,^41,42 these results support the hypothesis that compounds 2c, 2d, 3c, and 3d can be included in this category.

Motivated by these results, we evaluated the in vitro antioxidant activity of the same compounds tested in the aforementioned cytotoxic assay. We tested the ability to scavenge ABTS derived radicals and H₂O₂ oxidant. Ascorbic acid and Trolox (6-hydroxy-2,5,7,8-tetramethylchroman-2-carboxylic acid) were used as standard antioxidants in a comparison test. The assayed compounds potently inhibited ABTS radicals and H₂O₂, compared to the standards (Table 2).

Table 2. In Vitro Antioxidant Activity of Compounds 2c, 2d, 3c, and 3d.

	IC50 (μg/mL)a
Cmpd	ABTS	H2O2
2c	12.71 ± 0.25	15.89 ± 0.18
2d	14.26 ± 0.15	20.35 ± 0.27
3c	10.05 ± 0.09	18.56 ± 0.31
3d	9.43 ± 0.11	17.44 ± 0.18
Ascorbic Acid	12.75 ± 0.12	19.27 ± 0.54
Trolox	18.73 ± 0.26	20.38 ± 0.19

^aAll measurements were performed in triplicate.

Among the series, the dimethyl substituted compounds 3c and 3d exhibited a significant radical scavenging capacity on both ABTS and H₂O₂ with values of 10.05 and 18.56 μg/mL for 3c and 9.43 and 17.44 μg/mL for 3d, lower than compared standards, ascorbic acid and Trolox (12.75, 19.27 μg/mL and 18.73, 20.38 μg/mL, respectively).

Furthermore, in order to evaluate their drug likeness and the potential ability to cross the BBB, the compounds 2a–g and 3a–g were also in-silico scored for their physiochemical and pharmacokinetic parameters (ADME) by using the extended version of Lipinski’s rule of five. All the compounds were found to be BBB permeant, and none of them violate any Lipinski’s RO5 (Supporting Information).

In conclusion, we have synthesized a new series of ring-expanded diazepane-based compounds. The new series showed enhanced affinity than its original counterpart³⁵ toward both σR subtypes, and among the series, the benzofurane derivative 2c showed the best σ1R affinity and molecular dynamic simulations confirmed a strong interaction with the active site of the receptor. The benzofurane and quinoline derivatives 2c, 3c and 2d, 3d displayed the best K_iσ values and a safe profile toward two human cell lines. Altogether these data, along with the documented radical scavenging and cell survival promoting activities, support the interest for further studies aiming at evaluating the potential neuroprotective activity of this novel series of σ1R ligands.

Glossary

Abbreviations

σ1R and σ2R

sigma-1 and sigma-2 receptor

ABTS

2,2′-azino-bis(3-ethylbenzothiazoline-6-sulfonic acid)

ADME

absorption distribution metabolism excretion

ATB

automated topology builder

BBB

blood brain barrier

Boc

tert-butyloxycarbonyl

CNS

central nervous system

DCM

dichloromethane

DCVC

direct column vacuum chromatography

DTG

di-o-tolylguanidine

Et₃N

triethylamine

ER

endoplasmatic reticulum

DMEM

Dulbecco’s Modified Eagle’s Medium

MAM

mitochondrion-associated membrane

MD

molecular dynamics

MMPBSA

molecular mechanics Poisson–Boltzmann surface area

MW

molecular weight

NMDA

N-methyl-d-aspartic acid

NPC1

Niemann–Pick cholesterol transporter type1

RMSD

root-mean-square deviation

RMSF

root mean squared fluctuation

RO5

rule of five

TLC

thin-layer chromatography

TFA

trifluoroacetic acid

TMEM97

transmembrane protein-97

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsmedchemlett.9b00524.

• Detailed experimental, synthetic procedures and characterization of compounds, additional computational details, pharmacology and cytotoxicity assays, antioxidant activity, and drug likeness prediction (PDF)

Author Contributions

DZ synthesized and characterized the compounds and wrote the manuscript; SF generated the computational results and analysis; AC performed the antioxidant assay; MR and RM tested the cytotoxicity of the compounds and wrote the manuscript; DS and BW performed the binding assay; DZ and MGM conceived the project. The manuscript was written through contributions of all authors. All authors have given approval to the final version of the manuscript.

The authors declare no competing financial interest.