IMPURITY D OF ANTICANCER DRUG FULVESTRANT AS A POTENTIAL MULTIFUNCTIONAL INHIBITOR FOR THE MARBURG VIRUS

Haoqi Wang¹, Nirmitee Mulgaonkar¹, Samavath Mallawarachchi¹, Sandun Fernando^1*

ABSTRACT

A commonality between viruses and cancer is their ability to seize cellular machinery to rapidly proliferate while suppressing the host’s innate immune response. Using probe molecules, specific sites within the virion protein 35 (VP35) of the Marburg virus that have a high affinity to dsRNA (also known as the interferon inhibitory domain, IID), RNA-dependent RNA polymerase (RdRp), and nucleoside triphosphates (NTPs) were identified in silico. Using the dsRNA- binding site as the target, three potential drug molecules FID, CBSMA, and DOCHE were screened. We discovered that FID has a unique ability to simultaneously bind to all the three key binding domains with high affinity – a property that none of the other screened drug-like candidates possessed. The ability of FID to bind onto multiple domains of VP35 is significant since this provides the potential to thwart a viral infection from several fronts simultaneously using a single drug to combat rapidly mutating viruses.

Keywords: Marburg virus, VP35, Fulvestrant impurity D, Anticancer drug

Introduction

Marburg virus (MV), which causes Marburg virus disease (MVD), a form of viral hemorrhagic fever, is a Category-A bioagent [1, 2] that can cause unprecedented devastation. MVD was first discovered in 1967 when the importation of infected monkeys from Uganda caused epidemics in Germany and Yugoslavia [3]. Since then, several outbreaks of MVD have been reported, with the last outbreak occurring in 2017 in Kween district, Uganda [4-7]. MV is a filovirus - the same family that causes Ebola virus disease, and these viruses are among the most virulent human pathogens known; with the most recent outbreaks showing high pathogenicity and an extremely high human fatality rate between 23 to 90% [8, 9]. The main causes of fatality by MV include internal hemorrhage and organ failure [9-11]. Unfortunately, there are no licensed vaccines nor drugs available to treat an infection caused by the MV [12-14].

To confront MV, there are only a handful of proteins to target: NP (nucleoprotein), VP35 (polymerase cofactor), VP40 (matrix protein) [15], GP (glycoprotein), VP24 (minor matrix protein) [16], VP30 (transcription factor) [17], and L (RNA-dependent RNA polymerase [18], RdRp) and yet, no inhibitor has been uncovered that can effectively suppress viral replication. Moreover, in  MV, VP35 and VP40 proteins show immunosuppressive properties [19-21], adding another layer of defense from being attacked by our immune system. The VP35 present in both Marburg and Ebola viruses is a multifunctional protein that acts as a polymerase cofactor and a viral protein chaperone [19, 22, 23]. In addition, it can also act as an antagonist to the natural immune responses, including interferon (IFN) production and protein kinase R activation [24]. Due to its multifunctionality, it is believed that obstructing the RNA-dependent polymerase cofactor VP35 could be one of the most viable and yet unexplored approaches to silencing filovirus replication.

VP35 contains a central oligomerization domain with a predicted coiled-coil motif (residues 70-120 on sequence UniProtKB - P35259 (VP35_MABVM)) essential for RNA polymerase function and homo-oligomerization of the protein, a chaperoning peptide at the N-terminal that acts as a chaperone for viral proteins and an interferon (IFN) inhibitory domain (IID) at the C-terminal end (residues 204-329) that binds dsRNA [19]. VP35-mediated IFN antagonism correlates with double-stranded RNA (dsRNA) binding activity. The binding of VP35 IID to dsRNA can block the detection of dsRNA by immune sensors such as RGI-I and MDA5, thus inhibiting interferon production [19, 25]. Contrary to the Ebola virus VP35 IID, which has a higher affinity towards the blunt ends of dsRNA, MV IID tends to coat the backbone of dsRNA [26]. It is also able to inhibit phosphorylation of interferon regulatory factor 3 (IRF3) [27]. Since all these domains are critical for viral proliferation and virulence, the antiviral activity against MV could be achieved by targeting either of the domains above.

Isolating the RdRp binding domain of VP35 helps to understand the protein functions. RdRp catalyzes the transcription, capping, and polyadenylation of viral mRNAs. Specifically, the polymerase catalyzes negative single-stranded viral RNA ((-)ssRNA) replication. The template consisted of the viral RNA tightly encapsulated by the nucleoprotein (NP) . The viral RdRp binds the promoter region at the 3’ terminus of genomic RNA and proceeds with the transcription of viral mRNA. The viral phosphoprotein acts as a processivity factor. Capping of mRNA is concomitant with the initiation of mRNA transcription, and is done by GDP polyribonucleotidyl transferase (PRNTase) enzyme once the nascent RNA chain reaches a length of a few nucleotides. Ribose 2'-O-methylation of viral mRNA cap precedes and facilitates subsequent guanine-N-7 methylation; the viral polymerase carries both activities. Polyadenylation of mRNAs occurs by a stuttering mechanism at a slippery stop site present at the end of viral genes. Once the mRNA transcription is finished, the polymerase is able to resume transcription of the downstream gene . The reactions involved, and thus could be disrupted, by RdRp suppression are as follows [28]:

• Nucleoside triphosphate + RNA(n) à diphosphate + RNA(n+1)
• S-adenosyl-L-methionine + G(5')pppR-RNA à S-adenosyl-L-homocysteine + m₇G(5')pppR-RNA
• 5'-triphospho-mRNA + GDP à diphosphate + guanosine 5'-triphospho-mRNA
• S-adenosyl-L-methionine + a 5'-(N(7)-methyl 5'-triphosphoguanosine)-(2'-O-methyl-purine-ribonucleotide)-(ribonucleotide)-[mRNA] à S-adenosyl-L-homocysteine + a 5'-(N(7)-methyl 5'-triphosphoguanosine)-(2'-O-methyl-purine-ribonucleotide)-(2'-O-methyl-ribonucleotide)-[mRNA]

There is only a limited amount of research done on the discovery of drugs targeting MV VP35 [22]. One study reports the use of synthetic antibody sFab H3 as a specific inhibitor of MV IID [29], and some studies were done using non-human primates present siRNA and PMOplus as potential filovirus VP35 inhibitors [30]. However, these studies are still at the preliminary stage.

This work aims to identify inhibitors that have a high potential to impede viral replication and immune suppression via VP35 silencing by analyzing interactions between impinging ligands and the dynamically changing receptor via molecular dynamics (MD) simulations.

Materials and Methods

Protein sequence analysis

The VP35 protein in MV comprises of a central oligomerization domain (residues 70-120 on sequence UniProtKB - P35259 (VP35_MABVM)) essential for RNA polymerase function [19] and an interferon (IFN) inhibitory domain (IID) at the C-terminal end (residues 204-329) that binds dsRNA. VP35-mediated IFN antagonism correlates with the binding activity of double-stranded RNA (dsRNA). The central oligomerization domain has a known RNA polymerase function which is called the RNA binding domain (RBD). Experiments have shown that VP35 binds to dsRNA to avoid detection by the immune system. The crystal structure of the coiled-coil region of VP35 (PDB: 5TOH) [19] and RBD in MV VP35 bound with dsRNA (PDB: 4GHA) [26] obtained by X-ray diffraction was available in the Protein Data Bank. Residues 261 to 329 from this structure were predicted to interact with the RBD. Computational predictions of the three-dimensional structure of proteins were constructed using comparative homology modeling techniques. Since the crystal structure of the Marburg virus L-protein has not yet been solved, a homology model was developed using the amino-acid sequence UniProtKB - P31352 (L_MABVM) [31] to ascertain the binding behavior of VP35 on RdRp protein. The antiviral activity could be achieved by targeting either of the domains above.

RNA-dependent RNA polymerase (RdRp protein) interactions with VP35

The modeled RdRp protein and VP35 binding phenomena were resolved using the ZDOCK [32] server using the IID containing subunit (4GHA-chain A) and coiled-coil region (5TOH) of VP35 as the receptor and RdRp as the ligand and vice versa. ZDOCK is an interactive server that performs a rigid-body search for predicting docked conformations of two interacting proteins.

Double and single-stranded RNA and NTP interactions on VP35

RNA-interacting complexes were prepared by the default AutoDock Vina protocol, where RNA is considered as the ligand and VP35 as the receptor [33]. Two grid boxes were made based on the two known active sites of VP35, the IID, and RdRp binding pockets. Interaction diagrams and contacted amino acids were analyzed in MGLtools 4 [34].

The ligandability of NTP on VP35 was analyzed via AutoDock Vina [35]. The protocols were same as given above.

Druggability Assessment

Then, a druggability analysis [36] was done to ascertain the feasibility of using VP35 as a receptor that would respond to potential drugs by running 40 ns NAMD [37] simulations of the receptor in the presence of a solution containing small organic probe molecules. The probes, which are the representatives of typical drug-like molecules consisted of isopropanol (70%), acetamide (10%), isobutene (10%), isopropylamine (5%), and acetate (5%). The druggability of MV VP35 was assessed to locate the presence of individual or clusters of hot spots indicative of a potential drug binding site. The equilibrated system consisted of 2800 water and 140 probe molecules (i.e., 98 isopropanol, 14 acetamide, 14 isobutene, 07 isopropylamines, and 07 acetates,). Chloride ions were added to make the system charge neutral.

Pharmacophore Identification

Enhanced Ligand Exploration and Interaction Recognition Algorithm (ELIXIR-A) [38-41] (ELIXIR-A) was used for isolating pharmacophore points based on the analyses of interactions between VP35 and probe molecules. ELIXIR-A is an in-house pharmacophore screening algorithm that recognizes pharmacophoric features, i.e., the ensemble of steric, electrostatic, and hydrophobic properties that are crucial for optimum supramolecular interactions with the target receptor to inhibit its biological activity. The probe molecules were converted to pharmacophores using the ELIXIR-A VMD plugin. These pharmacophores were used for further compound screening.

Compound Screening and Verification

With the pharmacophore information obtained from ELIXIR-A, potential compounds were screened using the ZINCPharmer software [42] from the ZINC15 database [43]. Structure-guided pharmacophore-based screening focuses on identifying ligand conformations with potential pharmacophore features based on the functional groups (side chains of amino acids) present at the binding site of the receptor. The binding of molecules having a minimum of three pharmacophore points was validated in silico via AutoDock Vina using a molecular docking protocol. Vina uses a scoring function to evaluate several docked orientations for each small molecule and reports only the nine most stable conformations with the most negative binding score (i.e., the highest binding affinity). The binding of the compound having the highest affinity amongst those screened was further evaluated via MD simulations.

Intermolecular interaction analysis from MD simulations

The screened ligands were selected to study the possible mode of action (MOA) on the receptor after the docking simulation. The MD simulations were performed on the Schrödinger Desmond platform [44]. The system was solvated in an orthorhombic box using the Simple Point-Charge (SPC) solvent model with a buffer distance of 10 Å. The protein and ligand structures were optimized using Schrödinger's Protein Preparation Wizard [45]. The docked complexes were prepared by Glide [46]. The free N and C termini of the protein were capped for stabilizing the protein structure. All the missing hydrogens were added, and hydrogen bonds (H-bonds) were optimized. The strained minimization was performed with the OPLS3e force field [47]. Prior to simulation, each system was minimized using Schrödinger Desmond's default relaxation protocol. The simulations were performed under the OPLS3e force field. For equilibration, the system heavy atoms were first minimized with restrains under 10K, then increasing the temperature to 300K with restrains and the final relaxation step under 300K Normal Pressure and Temperature (NPT) ensemble. After relaxation, the simulations were carried out under the NPT ensemble at 300K and 1.01325 bar pressure for 20 ns. The recording interval was 20ps and 1000 frames were saved. Post simulation trajectory analyses were performed by Schrödinger Simulation Interactions Diagram (SID).

Results and Discussion

RNA-dependent RNA polymerase (RdRp protein) interactions with VP35

The interactions between RdRp and VP35 proteins were resolved using ZDOCK server (Figure 1). For ease of interpretation, we have kept the VP35 projection coordinates constant to the extent possible throughout the document.

Double and single-stranded RNA interactions on VP35The analysis generated several binding conformations of VP35 onto RdRp (Figure 1a). A simulation coiled-coil region of VP35 further confirmed the VP35 binding site on RdRp as depicted in Figure 1b with multiple conformations binding at the vicinity of residues 620-812 of RdRp as reported previously (Figure 1c). Simulations predicted two possible sites for RdRp binding on VP35 and based on further analysis the site demarcated in a blue circle spanning residues 210-299 was assigned as the RdRp binding domain on VP35 as depicted in Figure 1d.

To isolate RNA binding sites, a simulation of a double-stranded12-base-pair RNA (dsRNA) with RNA-dependent Polymerase MV VP35 protein (Lake Victoria strain Musoke 80; PDB: 4GHA) was performed.  This initial simulation predicted two sites that have a high affinity to dsRNA on VP35, consistent with that resulted from RdRp and VP35 simulations. After comparing the simulation output with the actual crystal structure of dsRNA-interfaced Marburg VP35 (PDB: 4GHA), the RdRp binding site and dsRNA binding sites were confirmed. The binding analysis suggests close interactions of the dsRNA with PHE228, ASN261, ARG271, PRO295, and LYS298. These results agree with previous experimental and simulation-based studies, which report PHE228, ARG271, ARG294, and LYS298 as the important residues for dsRNA binding, and mutations in these residues are reported to reduce dsRNA binding affinity [48]. The remaining site was confirmed to have strong interactions with RdRp (Figure 2) and was assigned to be the RdRp binding domain (RdRpD).

Since the ultimate objective of this analysis was to come up with potential viral inhibitors, it was decided to interrogate the protein with nucleotides of decreasing complexity. Accordingly, a docking simulation was done on VP35 initially with an 11-base ssRNA [35]. This analysis confirms the preferential binding of ssRNA to dsRNA site with a binding affinity of -13.1 kcal/mol while also confirming the binding ability of ssRNA to RdRpD with a lower, but significant affinity of -12.1 kcal/mol (Figure 2c), suggesting the susceptibility of both sites to RNA-analog-based inhibition. Obviously, ssRNAs are unlikely candidates as antiviral drugs owing to their large size; nevertheless, nucleotide triphosphate (NTP) and oligonucleoside triphosphate (OTP) analogs are promising candidates as target drugs. Accordingly, as the next step, how RNA NTPs interact with the VP35 was investigated. 

NAMD simulations with the four RNA NTPs revealed that the interaction space broadens significantly from two (dsRNA and RdRp binding sites in the case of double and single-stranded RNA) to three that included an additional NTP-binding domain (NTPD) (Figure 3). Of the NTPs, analysis of binding energies suggests that adenosine triphosphate (ATP) and uridine triphosphate (UTP) have a binding preference to dsRNA and RdRp-binding domains; however, all four nucleotides displayed a high affinity multi-conformational binding preference to NTPD at the vicinity of residues THR272, PHE273, and ASP274. The positioning of the NTPD on 4GHA with the coiled-coil region of VP35 (5TOI) suggests this site plays a role in routing the NTPs to the oligomerization domain of VP35 (Figure 1b).