The Study of Natural Compounds as Antidepressants by Bioinformatics Methods ^†

Abstract

According to the World Health Organization, neuropsychiatric disorders pose an increasingly greater burden on health, society, and economy of countries. Currently, there are strategies to prevent some of these disorders, and there are treatments or means to alleviate their symptoms. In the case of depression, the medication consists of antidepressant drugs. Such drugs can also be used in other conditions like anxiety, pain, or insomnia. A shortcoming of the currently used antidepressants is the occurrence of side effects that sometimes are unbearable for the patient. In this respect, a promising direction is the usage of medicinal plants. Plant parts are rich in phytochemicals that could be beneficial in mental disorders by acting on various targets. Here, we investigated the antidepressant effect of ten natural compounds from sage, mint, and citrus. The biological activity of these compounds was investigated by calculating pKi values and affinities for dopamine receptor D2, serotonin transporter (SERT), and 5-hydroxytryptamine receptor 1A (5-HT1A) using quantitative structure-activity relationship (QSAR) models. Our results showed that linalyl acetate, 1,8-cineole, and neryl acetate could be efficient antidepressants and neuroleptics.

1. Introduction

According to the World Health Organization (WHO), depression is a prevalent mental disorder that seriously affects the function of patients inside their family, at their workplace, or school [1]. Patients with mild to severe depression are offered treatment as antidepressant drugs and psychotherapy [2,3]. The drugs prescribed in depression fall into three categories: monoamine oxidase inhibitors, tricyclic antidepressants, and second-generation antidepressants, like norepinephrine, serotonin, or serotonin–norepinephrine reuptake inhibitors [4,5]. These present a series of adverse side effects like anxiety, tachycardia, tremor, sedation, blurred vision, etc. [4,6]. The intensity of these secondary effects is sometimes unbearable for patients that become intolerant to the treatment [7]. In this context, natural products might represent an alternative to conventional drugs for the treatment of depression.

Natural compounds may be a viable alternative in the treatment of viral and bacterial infections, as well as adjuvants and treatments for a variety of diseases [8,9,10]. Up to date, the beneficial effect of plant compounds in depression has been extensively reviewed. From these, we can mention Chinese herbal medicines [11], Ayurvedic single or holistic approaches [12], and many other phytochemicals or medicinal herbs [4,13]. The usage of natural compounds for therapy imposes some challenges concerning their pharmacokinetic and pharmacodynamic properties [14], meaning that a thorough characterization of these compounds is required. Cheminformatics methods can be of great assistance, allowing the description of the compounds and prediction of their biological effect [15,16].

Using the experience from our previous studies in predicting the anti-Alzheimer effect of natural compounds from Mentha spicata essential oil [14] or in predicting the antidepressant effect and the targets of candidate compounds [17,18], here, we used cheminformatics methods to investigate the antidepressant potential of ten natural compounds from diverse sources: 1,8-cineole (eucalyptus, sage), limonene (peppermint, spearmint), sabinene (lemon, mint), resveratrol (skin of grapes), chamazulene (german chamomile, roman chamomile), germacrene D (peppermint), linalyl acetate (sage), nerol (common grapes), neryl acetate (lemon balm, peppermint), and quercetin (grape). These compounds were mentioned in the literature as having antidepressant or neuroprotective effects [19]. We calculated the pharmacokinetic profiles of compounds. The prediction of their antidepressant effect was performed by considering three drug targets in depression, namely the serotonin transporter (SERT), dopamine receptor 2 (D2), and 5-hydroxytryptamine receptor 1A (5-HT1A). In the case of each target, we built a quantitative structure-activity relationship (QSAR) model that was further used to assess the effect of considered natural compounds [20].

2. Materials and Methods

2.1. Ligand Selection and Assessment of Their Drug and Lead-Likeness Features

We selected ten natural compounds from various vegetable sources, namely 1,8-cineole, limonene, sabinene, resveratrol, chamazulene, germacrene D, linalyl acetate, nerol, neryl acetate, and quercetin based on previous studies mentioning their possible beneficial effects in neuropsychiatric treatments. Calculations were performed using the Simplified Molecular Input Line Entry (SMILES) files of selected compounds that were fetched from the PubChem database [21]. The three-dimensional structures of compounds were modeled using Molecular Operating Environment (MOE) software (https://www.chemcomp.com/Products.htm, accessed on 30 May 2021) by applying the MMFF94x force field, Gasteiger-type charges, and a 0.005 gradient. The drug-likeness features of selected compounds were evaluated using Lipinski [22], Veber [23], Ghose [24], and Egan [25] filters implemented in the SwissADME web tool [26].

2.2. Computational Pharmacokinetics Profiles of Natural Compounds

The absorption, distribution, metabolism, excretion (ADME), and toxicity profiles of compounds were determined using the pkCSM database [27]. From all the properties calculated, we focused on the intestinal absorption (human, %), blood-brain barrier (BBB) permeability (log BBB), central nervous system permeability (CNS), fraction un-bound (human) and feature of renal organic cation transporter 2 (OCT2), AMES toxicity, hepatotoxicity, LD50 (median lethal dose) and maximum tolerated dose (human).

2.3. Predicted Biological Activities of Natural Compounds on SERT, 5-HT1A and D2 by 3D-ALMOND-QSAR

We used a non-aligned 3D-QSAR method and the 3D-QSAR-ALMOND method [17] to predict the biological activities of natural compounds at SERT, 5-HT1A, and D2 receptors. The activities were evaluated using Pentacle software (http://www.moldiscovery.com/software/pentacle/, accessed on 30 May 2021) and were expressed as pKi values (log 1/Ki, Ki represents the inhibition constant). We built three QSAR models as QSAR-SERT, QSAR-5-HT1A, QSAR-D2 with good statistical parameters r² fitted correlation coefficient (greater than 0.8) and q² cross-validated correlation coefficient (greater than 0.6). In building the models, we considered a synthetic antidepressant and neuroleptic molecules in the training and validation sets. PDSP Ki Database—Psychoactive Drug Screening Program [28] was used to retrieve the experimentally determined Ki values of molecules from training and validation sets. The test sets considered in the case of all three QSAR models included the selected natural compounds, namely 1,8-cineole, limonene, sabinene, resveratrol, chamazulene, germacrene D, linalyl acetate, nerol, neryl acetate, and quercetin.

3. Results

3.1. Drug-Likeness Features and Pharmacokinetics Profiles of Compounds

The application of Lipinski, Ghose, Veber, and Egan rules showed that all considered natural compounds present drug-like features. Predicted absorption, distribution, elimination, and toxicity features of compounds show that: (i) all compounds present a good intestinal absorption, (ii) all compounds except for quercetin present a good BBB permeability, (iii) the compounds except sabiene and chamazulene present a good CNS permeability, (iv) all compounds recorded a low human fraction unbound, (v) the compounds are not substrates of renal OCT2, and (vi) the compounds do not present toxicity features like hepatotoxicity, cardiotoxicity, or AMES toxicity.

3.2. Predicted Biological Activities of Natural Compounds on SERT, 5-HT1A and D2

Three 3D-QSAR-ALMOND models were built by considering the activity of molecules on each target (QSAR-SERT, QSAR-5HT-1A, QSAR-D2). Good statistical parameters were obtained when the QSAR models simultaneously considered the contribution of several descriptors like hydrophobicity, electrostatic, hydrogen bond donor/acceptor in various combinations. The QSAR models were used to predict the biological activities of selected natural compounds at SERT, 5-HT1A, and D2.

The biological activity of compounds at SERT was predicted using the QSAR-SERT model. We observed that seven natural compounds (limonene, sabiene, chamazulene, germacrene D, linalyl acetate, nerol, and neryl acetate) have a strong antidepressant character.

When predicting the biological activity of natural compounds at 5-HT1A using the QSAR-5-HT1A model, we observed that the natural compounds have a medium antidepressant activity.

In order to determine the neuroleptic activity of compounds, we applied the QSAR-D2 model. Results show that natural compounds like 1,8-cineole, linalyl acetate, neryl acetate, and quercetin have high biological activity on this target.

4. Discussions

4.1. The Most Active Natural Compounds on SERT, 5-HT1A and D2 Active Sites

The biological activities of the natural compounds considered here were computed as pKi, where a high pKi value indicates a high biological activity [29]. Prediction results were compared with the most active drug from the training series used to build the QSAR model. In the QSAR-SERT model, the most active natural compounds are linalyl acetate, chamazulene, neryl acetate, nerol, and germacrene D, as revealed by the comparison with the antidepressant drug paroxetine. The most active compounds in the QSAR-5-HT1A model are 1,8-cineole and linalyl acetate, as determined by comparison with ziprasidone, the most active compound in the QSAR-5-HT1A training set. Relative to spiperone, the most active compound in the QSAR-D2 training set, the most active natural compounds appear to be quercetin, neryl acetate, linalyl acetate, and 1,8-cineole. These results point toward linalyl acetate, 1,8-cineole, and neryl acetate as promising antidepressant and neuroleptic molecules. In Figure 1, we present their three-dimensional structures relative to the spatial structures of the drugs used as a reference, namely paroxetine, ziprasidone, and spiperone.

4.2. The Drug-Likeness Features and Pharmacokinetics Profiles of the Most Active Compounds

Drug-like filter validation is correlated with possible drug actions and good bioavailability. The most active natural compounds comply with all drug-likeness rules. Pharmacokinetic investigations reveal that the most active natural compounds are well tolerated by the human body. In

Table 1. Computed AMES, hepatotoxicity, hERG I and hERG II inhibitor for selected natural compounds and drugs.

Compound	Hepatotoxicity	AMES Toxicity	hERG I Inhibitor	hERG II Inhibitor
linalyl acetate	No	No	No	No
1,8-cineole	No	No	No	No
neryl acetate	No	No	No	No
paroxetine	Yes	Yes	No	Yes
ziprasidone	Yes	No	No	Yes
spiperone	Yes	No	No	Yes

, we present the evaluation of hepatotoxicity, cardiotoxicity, and AMES toxicity in the case of the most active natural compounds—neryl acetate, 1,8-cineole, and linalyl acetate. We also present the same parameters calculated in the case of the most active drugs from QSAR models training sets, namely paroxetine, spiperone, and ziprasidone.

According to our predictions, the natural compounds are not hepatotoxic, AMES toxic, or hERG I and II inhibitors. In contrast, regularly used medications—spiperone, paroxetine, and ziprasidone, all have hepatotoxicity and are hERG II inhibitors. Moreover, paroxetine is AMES hazardous (Table 1).

5. Conclusions

Here, we investigated the potential of ten natural compounds (1,8-cineole, limonene, sabinene, resveratrol, chamazulene, germacrene D, linalyl acetate, nerol, neryl acetate, and quercetin) to exhibit antidepressant effects by acting on SERT and 5-HT1A receptors or neuroleptic activity acting on D2 receptors. This was performed using three QSAR models developed for predicting the activity of compounds at each target.

Before building the models, the compounds were filtered by drug and lead-likeness rules showing that all compounds comply with the rules, presenting drug-likeness and good bioavailability features. Moreover, the compounds were filtered based on their predicted ADME and toxicity profiles and all compounds resulted in having good BBB and CNS permeability while being non-toxic on the liver or heart and not mutagenic.

By applying the QSAR models that we built, we noticed that seven out of ten compounds should have good biological activity on SERT, two compounds should modulate 5-HT1A receptors, and four compounds should modulate D2 receptors. From these, linalyl acetate appears the only compound modulating all three protein targets, 1,8-cineole should modulate 5-HT1A and D2 receptors, and neryl acetate should modulate SERT and D2 receptors.