Pharmaceuticals for Targeting Coronaviruses

Structure-Based Drug Discovery Approaches Applied to SARS-CoV-2 (COVID-19)

Igor José dos Santos Nascimento¹, Thiago Mendonça de Aquino¹, Edeildo Ferreira da Silva-Júnior¹, ², *

¹ Chemistry and Biotechnology Institute, Federal University of Alagoas, Maceió, Brazil

² Laboratory of Medicinal Chemistry, Pharmaceutical Sciences Institute, Federal University of Alagoas, Maceió, Brazil

Abstract

Viral diseases have caused millions of deaths around the world. In the past, health organizations and pharmaceutical industries have neglected these diseases for years, mainly because they affected a small geographic population. In contrast, since 2016, several viral outbreaks have been reported worldwide, such as those caused by Ebola, Zika, and SARS-CoV2 (COVID-19). Thus, these have received more attention, leading to increased efforts to search for new antiviral drugs. The SARS-CoV-2 pandemic, already responsible for more than 1,254,567 deaths worldwide, is the greatest example of a virus that has always been present in our society, responsible for small outbreaks in Asian and Arabic countries in 2004 and 2012. But, investments in research to identify/discover new drugs and vaccines were only intensified in 2020, in which only the remdesivir (an FDA-approved drug) was developed to addressCOVID-19 until today. Nonetheless, it has been used in hospitals in the United States and Japan, in emergency cases. Indeed, it justifies greater investments in discovering new alternatives that could save thousands of people. In this context, improving drug discovery techniques is fundamental in searching for new therapies that could be selective and effective to combat SARS-CoV-2. Drug discovery approaches are based on ligands (Ligand-Based Drug Design - LBDD) or structures (Structure-Based Drug Discovery - SBDD). Concerning SBDD, it is the main and most evolved technique used for discovering new drugs. The application of SBDD techniques has improved the pharmacological arsenal against diverse diseases, which allowed the discovery of innovative treatments, such as inhibitors of HIV-1 proteases. In this chapter, main SBDD techniques (i.e. homology modeling; molecular dynamics and docking; de novo drug discovery; pharmacophore modeling; fragment-based drug discovery; and virtual high-throughput screenings) applied to discover new hit compounds SARS-CoV-2 (COVID-19) will be discussed in details.

Keywords: Drug discovery, Dynamics simulations, Molecular modeling, SARS-CoV-2, Structure-Based Drug Discovery, TMPRSS2, Virtual screening.

* Corresponding author Edeildo Ferreira da Silva-Júnior: Chemistry and Biotechnology Institute, Federal University of Alagoas, Maceió, Brazil and Laboratory of Medicinal Chemistry, Pharmaceutical Sciences Institute, Federal University of Alagoas, Maceió, Brazil; Tel: +55-87-9-9610-8311; E-mail: [email protected]

1. Introduction

On December 31st, 2019, an outbreak of pneumonia was reported caused by an unknown etiologic agent in Wuhan, a province of Hubei in China. Thus, with the sporadic number of cases, on January 9th, 2020, the new coronavirus was recognized as the causative agent by the Chinese Center for Disease Control and Prevention (CDC). When it started spreading at an alarming pace to other countries in the world, the new coronavirus (SARS-CoV-2, or COVID-19) was declared a pandemic by the world health organization (WHO) on March 11th, 2020 [1-3].

Since its discovery, SARS-CoV-2 has been responsible for several victims worldwide. To date (09/11/2020), there are already 50,266,033 confirmed cases with 1,254,567 deaths [4]. The main symptoms are fever, cough, fatigue, myalgia, and dyspnea. Its transmission occurs mainly through coughing, sneezing, and respiratory droplets [5]. These alarming statistics make research groups from around the world focus on discovering new therapies against this pandemic virus [6]. Advances in drug developments resulted in the repurposing of remdesivir in the United States. However, this drug still does not show the best effectiveness. So, a molecule that could be effective in eliminating SARS-CoV2 from the body is an unmet needed [6, 7].

Currently, biological targets guide the process of discovering new drugs. Then, the structure of a macromolecule is fundamental for this process [8]. Such structures provide valuable information on mechanisms of action and their correlation with biological activity [9]. In addition, information about the biological target and the availability of three-dimensional structures for these therapeutically attractive targets have resulted in several advances in the identification of inhibitors, as well as potential binding sites, contributing to the basis of structure-based drug discovery strategies (SBDD) [10].

In addition, in silico methods are increasingly gaining more visibility in the drug development field. These methods are used in SBDD and are related to higher chances of success with less financial cost and less time-consuming [11, 12].

In this context, this chapter will be addressed to the main SBDD techniques (homology modeling; molecular docking and dynamic; pharmacophore modeling; virtual screening and virtual high-throughput screening; fragment-based drug design; and de novo drug design) applied for the discovery of new promising compounds against SARS-CoV2.

2. Coronaviruses: History and Structure

Coronaviridae is a family of several groups of viruses responsible for the infection of both animals and humans. From this family, there are seven viruses that can infect humans, being: Severe Acute Respiratory Syndrome Coronavirus (SARS-CoV); Middle East Respiratory Syndrome Coronavirus (MERS-CoV); Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2); HCoV-OC43; HCoV-KHU1; HCoV-NL63; HCoV-229E [13]. Among these, the first three belong to the genus Betacoronavirus, and all of them display high potential for mutability, leading to plasticity and genetic variability, which contributes to their adaptation to different types of hosts [13, 14].

The first discovery of SARS-CoV was around the 1960s [15]. This pathogen is related to flu-like symptoms. However, its progress generates respiratory failure and, in many cases, death since it presents a higher mortality rate [16]. The SARS-CoV is a virus from animal reservoirs (in this case, bats) that can spread to other animals and humans, initially reported in Guangdong (China), in 2002 [16-18]. One year later, it spread to Asia and America, affecting 26 nations and causing 8,000 deaths. After its control, other reports were associated with laboratory accidents or transmission from animals to humans [16].

Concerning MERS-CoV, the main reservoir is dromedary camels and bats. These pathogens can infect bat cells through the receptor dipeptidyl peptidase-4 (DPP4), which is similar to the human receptor. The MERS-CoV exhibited widespread exposure in the Middle East and North, as well as in East Africa [17]. This disease was initially identified in 2012, in Saudi Arabia, and it has spread to about 20 countries. Since then, MERS-CoV has been detected in Europe, the Gulf region, and Korea. Since 2016, it is estimated that were infected approximately 1,638 people, of which around 35% were fatal victims [16].

As mentioned in the introduction, a new CoV variant was detected in Wuhan, China (December 2019), giving rise to one of the most significant outbreaks of unknown viral pneumonia. The new SARS-CoV (SARS-CoV-2) is more genetically similar to the SARS-CoV than MERS-CoV. Thus, it could be used information from SARS- and also MERS-CoV to discover new therapies [19, 20].

Deeming the knowledge about the structure and function of the virus, it is possible to model drugs with a focus on each target. In this context, the structure of SARS-CoV-2 is composed of structural and non-structural proteins, being used frequently for the design of new inhibitors. The structural proteins are spike (S) glycoprotein, membrane (M), envelope (E), and nucleocapsid (N) proteins. Among these, S protein is one of the most promising targets in drug discovery for SARS-CoV-2. This protein is related to viral entry by recognition of the membrane receptor and membrane fusion, mainly interacting with the Angiotensin-Converting Enzyme-2 (ACE-2) [21, 22]. In Fig. (1), it is shown the structural proteins from the SARS-CoV-2.

Fig. (1))

Structural proteins from SAR-CoV-2.

In total, SARS-CoV-2 has 16 non-structural proteins with different functions. These proteins are the main protease (3CLpro or CLpro or Mpro or nsp5) papain-like protease (PLpro ou nsp3), RNA-dependent RNA polymerase (RdRp ou nsp12), complex nsp7_nsp8, methyltransferase stimulating factor complex nsp16_nsp10; complex nspP10_nsp16, binding proteins nsp9; and endoribonuclease nsp15 [21]. Among these, the 3CLpro seems to be the most attractive target for drug discovery against the SARS-CoV-2 since it is responsible for cleaving polypeptide sequences after the glutamine residue. Moreover, there are no human proteases with similar structures or functions, making 3CLpro an ideal target for designing new drugs [23].

3. Drug Discovery Process

Developing a new drug is a costly and time-consuming procedure [24, 25]. The estimated time to discover a drug is about 12-14 years, costing approximately US$ 1 billion [26, 27]. Before the substance reaches clinical trials, several steps are needed, which include evaluation and its effectiveness, adverse effects, pharmacokinetics, and other parameters. Additionally, combinatorial chemistry and high-throughput screening (HTS) has become quite common in drug development groups. However, there is a need for methods that reduce the financial cost related to research, making drug discovery enter in "big data era", which refers to a large amount of data using mainly the field of information technology [24].

The drug discovery process is divided into 4 stages being 1) Selection and validation of the biological target; 2) Screening of compounds in a database and lead optimization; 3) preclinical studies; and 4) Clinical studies. In this context, in silico studies are widely used in steps one and two, in order to decrease the number of candidates in biological tests and increase the possibility of obtaining new hits [28].

In this context, computer-aided drug design (CADD) methods emerged to reduce the time (approximately 50%) and costs associated with the search for new therapeutic agents. Two approaches are more common within CADD methods, namely SBDD and ligand-based drug design (LBDD) [26, 29].

SBDD is a strategy based on information about the 3D-structures of targets Thus, the 3D-structure normally refers to the crystalline structure of a target complexed with a ligand. These structures can be used in the screening of large libraries of compounds. The screening can be rationally guided, showing the ligand’s complementarity at the binding site, improving the potency and selectivity of molecules [26, 30]. Molecular docking is the main technique used in SBDD protocols, which provides conformations and interactions of a ligand with a macromolecule. Pharmacophore models are also widely used to guide virtual screenings and designing new active ligands [30]. In cases where there are no 3D-structures for the target, homology modeling can be used to solve this problem by building a target from 3D-based template with high similarity via alignment with other targets (from the same organism or other organisms) [30].

LBDD strategy is usually applied when there is no crystal structure for targets. In sense, LBDD is performed based on ligands, which represent a set of inhibitors with well-known activity. Still, utilizing pharmacophore groups, the analysis of the similarity between ligands, or the development of Quantitative Structure-Activity Relationship (QSAR) models, results in successful strategies in drug design [26, 30]. In Fig. (2), is presented the main techniques comprised of LBDD and SBDD approaches.

Fig. (2))

Main techniques used in CADD.

4. SBDD Strategies and Discovery of Hits Against Coronaviruses

4.1. Homology Modeling

The main factor related to success in an SBDD approach is to obtain the 3D-structure of the targeted macromolecule [31]. The main database for obtaining 3D-protein structures is the Research Collaboratory for Structural Bioinformatics Protein Data Bank (RCSB PDB), which contains thousands of experimentally obtained structures. Additionally, the National Center for Biotechnology Information (NCBI) database is one of the main servers for obtaining protein sequences and performing amino acids’ alignments and contains millions of deposited sequences. Furthermore, it is clear that the number of deposited sequences is higher than 3D-structures’ number [32].

Given the fact that there are more sequences deposited than 3D-structures of proteins, the prediction of these structures made through computational techniques has been considered promising. Thus, when there is no a 3D-model of a targeted protein, homology modeling can fill that gap by using its amino acid sequence (normally in FASTA format). Therefore, homologous protein sequences with known 3D-structures models are used and thus generate a 3D-structure for the intended target, in which it has not been experimentally characterized [33, 34].

For a model to be considered homologous should present over a 30% similarity index between the template and retrieved sequences. A convenient homology procedure is based on the following steps: 1) Identification of a known 3D-structure that could serve as a model for building a hypothetical model; 2) Alignment of the template and model proteins’ sequences; 3) Building 3D-models from the alignments; 4) Validation of the best built homologous model. For the refinement, these steps should be repeated as many times as necessary until obtaining the ideal model validated [35] (Fig. 3). All these steps can be performed using the SWISS-MODEL web server, which is one of the most used tools for building homology models [36].

Fig. (3))

Typical homology modeling procedures.

A study performed by Dong and colleagues (2020) was carried out to build homologous models using amino acid sequences of structural and non-structural proteins from SARS-CoV2 obtained at the NCBI. They used the BLAST server to identify the best models and perform homology modeling [37]. As a result, it was demonstrated that the ORS1ab protein from SARS-CoV2 is highly similar to that one from SARS- and MERS-CoV, with similarity indexes of 90% and 60%, respectively. Moreover, the authors showed that non-structural proteins could be built using homology models, such as 3CLpro, which exhibited a similarity index of 94%, in comparison with known templates. Also, it was verified that the structural S, E, and N proteins could be built using the SARS-CoV structures as a template. Finally, this study may be useful for the virtual screening of new compounds against SARS-CoV-2. Similarly, Grifoni et al (2020) showed that SARS-CoV is the most similar to SARS-CoV-2 in phylogenetic and sequence identity analyses. By bioinformatic techniques, it was possible to identify B and T cell epitopes from SARS-CoV-2 that could be effectively recognized by the human immune response and thus could be promising targets for discovering new vaccines against this virus [38]. Regarding the identification of epitopes, Tilocca et al. (2020) carried out a homology study involving the main epitopes from coronaviruses N proteins [39]. Then, they showed a high-similarity index between SARS-CoV-2 N protein and RaTG13 (99%), while for SARS-CoV vs SARS-CoV-2 N protein was 90%; and 88% for SARS-CoV-2 vs pangolin. Also, epitope mapping by homology showed a potential immunogenic value in low identity sequences with SARS-CoV-2 N proteins. Finally, these observations may help in the discovery of new drugs for the treatment and prevention against SARS-CoV-2.

Still focused on demonstrating the similarity between SARS-CoVs, Uddin and colleagues (2020) carried out homology studies to investigate the origin of SARS-CoV-2, as well as similarities between its structural proteins [40]. Thus, there was a high similarity index between these SARS-CoV-2 proteins with those from SARS-CoV, in which S, N, M, and E proteins showed 36-95% coverage and similarity indexes ranging from 40 to 90%.

Bai and coworkers (2020) performed calculations to determine binding free energy values from SARS-CoV-2 and SARS-CoV during their interaction with the ACE2 receptor [41]. Initially, it was necessary to build a model by homology modeling. The authors have built models for ACE2/SARS-CoV-2 and m396 (antybody)/SAR-CoV-2 complexes. They showed that residues outside the binding domain were the main ones related to the most potent binding with SARS-CoV-2. The essential SARS-CoV-2 evolution occurs in the binding domain from the trimeric body of S protein, which facilitates conformational alterations and infection process by virus binding in ACE2. In addition, its connection with the m396 antibody shows the lowest energy contribution, which explains the lack of cross-reactivity with the antibody.

Wu and colleagues (2020) performed a homology modeling study of proteins encoded by SARS-CoV-2 against the proteins from other coronaviruses. Also, they built 19 homology models and performed their virtual screening in three databases, being FDA-approved drugs from the ZINC; compounds from traditional Chinese medicine; and 78 antiviral compounds commonly used in virtual screening for coronaviruses [42]. Still, human proteins ACE2 and TMPRSS2 were also built by homology modeling and used in their virtual screening protocol. Thus, the authors proposed several compounds that could be experimentally screened in order to verify their effectiveness against SARS-CoV-2, comprising antivirals (valganciclovir (1), ribavirin (2), and thymidine (3)); antimicrobials (phenethicillin (4), oxytetracycline (5), cefpiramide (6), sulfasalazine (7), doxycycline (8), demeclocycline (9), lymecycline (10), and tigecycline (11)); antiasthmatics (fenoterol (12), montelukast (13), and reproterol (14)), and among others (Fig. 4). Additionally, the authors proposed that Nsp3b, Nsp3c, Nsp7-Nsp8 complex, Nsp14, Nsp15, 3CLpro, PLpro, E-channel, RdRp, helicase, ACE2, and S proteins are the most favorable targets for these drugs. Finally, they demonstrated that remdesivir triphosphate binds strongly to RdRp, inhibiting RNA synthesis, with affinity energy of -112.8 kcal/mol. In addition, it binds strongly to TMPRSS2. The drugs lopinavir/ritonavir had no affinity for the 3CLpro, PLpro, RdRp, and others. Thus, the authors suggested that these drugs are not suitable to treat SARS-CoV-2 symptoms. In conclusion, through homology and virtual screening methods, the authors provided essential information for repurposing drugs for SARS-CoV-2.

Fig. (4))

Chemical structures of drugs identified by Wu and colleagues (2020) as promising compounds against SARS-CoV-2.

Hall and coworkers (2020) used homology modeling to build SARS-CoV S protein, in order to perform a virtual screening using 3,447 FDA-approved drugs [43]. They built a 3D-structure of the intended target by using the S protein from the SARS-CoV as a sequence template. Docking studies of the compound (15) (-7,234 kcal/mol) towards S protein and (16) (-11,016 kcal/mol) (Fig. 5) towards 3CLpro.

Fig. (5))

Chemical structures of compounds identified by Hall and coworkers (2020).

Similar to previous studies, Feng and colleagues (2020) performed a homology study by using structural proteins from SARS- and SARS-CoV-2. In addition, the authors performed a virtual screening in a library of 1,234 FDA-approved drugs towards S protein [44]. There was observed a high similarity between all proteins in this study. Then, the screening was carried out with 13 hits (dactinomycin (17), glycyrrhizic acid (18), eltrombopag (19), azilsartan medoxomil (20), bictegravir (21), temsirolimus (22), dolutegravir (23), elbasvir (24), irbesartan (25), gliquidone (26), tasosartan (27) lanreotide (28), and velpatasvir (29)) (Fig. 6), displaying affinity energies ranging from -9.3 to -12.3 kcal/mol, and interactions with the main amino acid residues, as well.

4.2. Pharmacophore Modeling

A pharmacophore group shows molecular characteristics or structural elements responsible for the biological activity of specific molecules [45, 46]. This term has gained more prominence in recent years since it is related to the discovery of new drugs, being useful in the pharmacophore-based virtual screening protocols for identifying hits and leads compounds [45].

Although the concept of pharmacophore is older than the discovery of computers, it has become essential in CADD, including mainly the molecular docking technique [45, 47]. Each atom has its characteristics of molecular recognition, such as H-bond donors or