Vitamin D3 and its hydroxyderivatives as promising drugs against COVID-19: a computational study

Abstract

The epidemiologic correlation between the poor prognosis of SARS-CoV-2 infection and vitamin D deficiency has been observed worldwide, however, their molecular mechanisms are not fully understood. In this study, we used combined molecular docking, molecular dynamics simulations and binding free energy analyses to investigate the potentials of vitamin D3 and its hydroxyderivatives as TMPRSS2 inhibitor and to inhibit the SARS-CoV-2 receptor binding domain (RBD) binding to angiotensin-converting enzyme 2 (ACE2), as well as to unveil molecular and structural basis of 1,25(OH)₂D3 capability to inhibit ACE2 and SARS-CoV-2 RBD interactions. The results show that vitamin D3 and its hydroxyderivatives are favorable to bind active site of TMPRSS2 and the binding site(s) between ACE2 and SARS-CoV2-RBD, which indicate that vitamin D3 and its biologically active hydroxyderivatives can serve as TMPRSS2 inhibitor and can inhibit ACE2 binding of SARS-CoV-2 RBD to prevent SARS-CoV-2 entry. Interaction of 1,25(OH)₂D3 with SARS-CoV-2 RBD and ACE2 resulted in the conformation and dynamical motion changes of the binding surfaces between SARS-CoV-2 RBD and ACE2 to interrupt the binding of SARS-CoV-2 RBD with ACE2. The interaction of 1,25(OH)₂D3 with TMPRSS2 also caused the conformational and dynamical motion changes of TMPRSS2, which could affect TMPRSS2 to prime SARS-CoV-2 spike proteins. Our results propose that vitamin D3 and its biologically active hydroxyderivatives are promising drugs or adjuvants in the treatment of COVID-19.

Introduction

The COVID-19 pandemic caused by SARS-CoV-2 presents a great threat to public health (Dong et al., 2020). One important epidemiologic correlation with SARS-CoV-2 infection (Alipio, 2020; Rhodes et al., 2021) is the prevalence of vitamin D deficiency in patients with severe COVID-19 symptoms (D’Avolio et al., 2020; Ebadi & Montano-Loza, 2020; Grant et al., 2020; Jakovac, 2020; Meltzer et al., 2020; Merzon et al., 2020; Rhodes et al., 2021; Slominski et al., 2020b, 2020c). This observation is not limited to a region or a particular type of population, instead is noticeable worldwide (Lau et al., 2020; Rhodes et al., 2021). Vitamin D belongs to a group of fat-soluble secosteroids of which enzymatic activation leads to the production of hormonally active hydroxy-metabolites (Jenkinson, 2019; Tuckey et al., 2019) that play a central role in body regulation of calcium, magnesium, and phosphate homeostasis and which control the expression of roughly 5% of human genes (Bikle & Christakos, 2020; Holick, 2003; 2007), and regulate cell proliferation and differentiation, immune-regulation, interaction with viral factors, autophagy and apoptosis (Aygun, 2020; Chen et al., 2020; Samuel & Sitrin, 2008; Slominski et al., 2014b, 2018; Zhang et al., 2020b; Zmijewski & Carlberg, 2020). In the canonical pathway, vitamin D3 is activated through two sequential hydroxylations at position C25 by CYP2R1 and CP27A1 with following hydroxylation at C1α by CYP27B1 to produce biologically active 1,25(OH)₂D3 (Bikle & Christakos, 2020; Holick, 2003). 1,25(OH)₂D3 regulates calcium body homeostasis and has important pleiotropic effects on multiple body functions including immune system (Bikle, 2008, 2010a, 2010b, 2011; Holick, 2003, 2007; Plum & DeLuca, 2010). An alternative, non-canonical pathway of vitamin D activation is initiated by series of hydroxylation of its side chain by CYP11A1 (Slominski et al., 2015, 2020a) with additional hydroxylations by other CYP enzymes (Slominski et al., 2020a; Tuckey et al., 2019). Metabolites of this alternative pathway are produced in vivo and are detectable in the human body (Slominski et al., 2012, 2015). They are biologically active with the mechanism of action being both different and overlapping with classical 1,25(OH)₂D3 (Slominski et al., 2014a, 2014b, 2017, 2018, 2020a). Of interest, CYP11A1 is also expressed in immune cells in a context dependent fashion (Slominski et al., 2020d). Using high doses of vitamin D3 and its hydroxyderivatives for SARS-CoV-2 prevention and therapy has been proposed (Entrenas Castillo et al., 2020; Grant et al., 2020; Kaufman et al., 2020; Muralidharan Jothimani et al., 2021; Quesada-Gomez et al., 2020; Slominski et al., 2020b, 2020c) but the exact mechanism of this action remains unknown.

Acute respiratory distress syndrome (ARDS) develops because of SARS-CoV-2 infection causing multiple organ damage in patients (Coronaviridae Study Group of the International Committee on Taxonomy of Viruses, 2020; Huang et al., 2020; Mehta et al., 2020). SARS-CoV-2 patients have increased immune responses with high concentrations of proinflammatory interleukins (ILs), tumor necrosis factor-α (TNF-α), gamma interferon (IFN-γ) and nuclear factor-κB (NFκB) (Conti et al., 2020; Huang et al., 2020; Zhang et al., 2020a) – leading to a phenomenon called the cytokine storm caused by triggered Th1 cell response (Cantorna & Mahon, 2005; Conti et al., 2020; Coronaviridae Study Group of the International Committee on Taxonomy of Viruses, 2020; Huang et al., 2020; Mehta et al., 2020; Schleithoff et al., 2006; Zhang et al., 2020a). 1,25(OH)₂D3 and CYP11α1-derived vitamin D metabolites decrease Th1 cell responses and IL-17 production (Ardizzone et al., 2009; Bikle & Christakos, 2020; Cantorna & Mahon, 2005; Chen et al., 2013; Janjetovic et al., 2009; 2010; Provvedini et al., 1983; Schleithoff et al., 2006; Slominski et al., 2014a, 2014b; Sun et al., 2018; Tsoukas et al., 1984). NFκB activity and TNF-α, IFN-γ and IL-2 production are also reduced by active forms of vitamin D.

The cellular entry of SARS-CoV-2 involves binding of receptor binding domain (RBD) of spike protein in SARS-CoV2 with angiotensin-converting enzyme 2 (ACE2) receptor, and cellular serine protease TMPRSS2 primes viral spike proteins in SARS-CoV-2 (Hoffmann et al., 2020). SARS-CoV-2 spike protein has lower binding affinity to ACE2 compared to SARS-CoV spike protein, but SARS-CoV-2 RBD has higher ACE2 binding affinity than SARS-CoV RBD (Perrotta et al., 2020). SARS-CoV and SARS-CoV-2 RBD are typically in standing up state and resting state, respectively. Although SARS-CoV-2 RBD has higher binding affinity, its lying-down state makes it less accessible to ACE2 and other inhibitory or neutralizing agents. To overcome this and maintain its high binding affinity, SARS-CoV-2 RBD has developed a strategy which involves the use of host protease for its activation. Furin, along with TMPRSS2 and cathepsins, have roles in viral entry of human cells (Perrotta et al., 2020; Shang et al., 2020). The hidden RBD of SARS-CoV-2 is critical to both vaccination and antibody neutralization due to its limited accessibility (Perrotta et al., 2020; Shang et al., 2020). TMPRSS2 inhibitors approved for clinical use blocked viral entry (Hoffmann et al., 2020). Molecules that would inhibit the interactions of SARS-CoV-2 RBD and ACE2 and would serve as TMPRSS2 inhibitors could be an effective therapy to constrain cellular entry of SARS-CoV-2.

In this study, we aim to determine the potential of vitamin D3 and its hydroxyderivatives to act as TMPRSS2 inhibitors and to inhibit ACE2 and SARS-CoV-2 RBD interaction using molecular docking, molecular dynamics (MD) simulations and binding free energy analyses. The results suggest that vitamin D3 and its hydroxyderivatives can serve as promising therapeutics or adjuvants against COVID-19.

Materials and methods

Protein receptors and ligands

ACE2 and SARS-CoV-2-RBD structures and the binding sites between ACE2 and SARS-CoV-2-RBD

Electron microscopy structure of ACE2 and SARS-CoV-2-RBD complex (PDB ID: 6M17, resolution: 2.90 Å) (Yan et al., 2020) was obtained from protein data bank (https://www.rcsb.org/). Monomer structures of ACE2 and SARS-CoV-2-RBD were used for determining if and how vitamin D3 and its derivatives interact with ACE2 or SARS-CoV-2-RBD. The binding sites between ACE2 and SARS-CoV-2-RBD were determined using contact amino acid function in PyMOL based on the complex structure of ACE2 and SARS-CoV-2-RBD. The binding sites between ACE2 and RBD of spike protein primarily occur on the N-terminal alpha helix of ACE2 and the arch at the top of the RBD of spike (Yan et al., 2020). Specifically, the residues Tyr41, Gln42, Lys³⁵³, Arg³⁵⁷, Asp³⁰, His34, Gln24 and Met82 from ACE2, and the residues Gln⁴⁹⁸, Thr⁵⁰⁰, Asn⁵⁰¹, Lys⁴¹⁷, Tyr⁴⁵³, Gln⁴⁷⁴ and Phe⁴⁸⁶ from SARS-CoV-2-RBD have been shown to interact (Yan et al., 2020). This binding site was used for molecular docking of vitamin D3 and its hydroxyderivatives with ACE2 and SARS-CoV2-RBD (red and blue region in Figure 1, respectively). The contact surface between ACE2 and SARS-CoV-2-RBD is shown in Figure 1.

Figure 1.

Binding surfaces between ACE2 and SARS-CoV-2-RBD. Contact surface of ACE2 shown as red and contact surface of SARS-coV-2 RBD shown as blue.

TMPRSS2 structure

TMPRSS2 structure is not available in protein data bank. Homologous modeling was used with MODELLER (Eswar et al., 2006) to predict the structure of TMPRSS2 based on the crystal structure of TMPRSS1 (PDBID: 1Z8G (Herter et al., 2005), resolution: 1.55 Å), a homologous protein of TMPRSS2 that has a 51.3% sequence similarity. TMPRSS2 model construction and evaluation were detailed in Supplemental Materials. The active site of TMPRSS2 is composed of a catalytic triad, consisting of His²⁹⁶, Asp³⁴⁵ and Ser⁴⁴¹, and a substrate binding residue Asp⁴³⁵, at the bottom of the pocket (Meng et al., 2020; Wilson et al., 2005). This active site was used for molecular docking of vitamin D3 and its hydroxyderivatives with TMPRSS2.

Vitamin D3 and its hydroxyderivatives

Vitamin D3 and its hydroxyderivatives including 1α,25(OH)₂D3, 25(OH)D3, 20S(OH)D3, 1α,20S(OH)₂D3, 20S,23S(OH)₂D3, 1α,20S,23S(OH)₂D3, 20S,23R(OH)₂D3 are the ligands with their structures shown in Table 1. They are used in this study to evaluate their potential as TMPRSS2 inhibitor and to inhibit ACE2 and SARS-CoV-2 RBD interactions.

Table 1.

Molecular docking and MM-PBSA binding free energy results for the complexes of vitamin D3 and its hydroxyderivatives with ACE2, SARS-CoV2-RBD and TMPRSS2 (kcal/mol).

Compounds╲Protein receptor		Binding affinity from molecular docking (kcal/mol)			MM-PBSA binding free energy (kcal/mol)a
		TMPRSS2	ACE2	SARS-CoV-2 RBD	TMPRSS2	ACE2	SARS-CoV-2 RBD
Vitamin D3		−6.3	−5.7	−6.0	−16.58	−17.60	−20.33
1α,25(OH)2D3		−6.6	−5.9	−4.6	−12.97	−30.45	−22.39
25(OH)D3		−6.4	−6.1	−6.3	−14.38	−25.35	−24.18
20S(OH)D3		−6.3	−5.9	−3.1	−17.16	−19.25	−4.44
1α,20S(OH)2D3		−6.0	−5.7	−6.6	−16.14	−14.92	−16.02
20S,23R(OH)2D3		−6.0	−5.8	−6.0	−12.75	−3.34	−11.36
20S,23S(OH)2D3		−5.4	−7.7	−6.0	−10.37	−25.11	−15.95
1α,20S,23S(OH)3D3		−6.0	−7.6	−5.5	−27.74	−12.75	−14.60

^aMM-PBSA binding free energy was calculated based on the top-ranked docking complex, in which entropy was not included. Molecular docking binding affinity < −5.0 kcal/mol and MM-PBSA < −10 kcal/mol are considered favorable binding.

Molecular docking of vitamin D3 and its hydroxyderivatives with ACE2, SARS-CoV2-RBD and TMPRSS2

We performed the molecular docking using AutoDock Vina software (Trott & Olson, 2010). Vitamin D3 and its hydroxyderivatives (1α,25(OH)₂D3, 25(OH)D3, 20S(OH)D3, 1α,20S(OH)₂D3, 20S,23S(OH)₂D3, 1α,20S,23S(OH)₂D3, 20S,23R(OH)₂D3) are the ligands, and ACE2, SARS-CoV2-RBD and TMPRSS2 were receptors for molecular docking. Autodock vina is supported by Autodock tools program (Morris et al., 2009) that was used for protein optimization by removing water and other molecules and adding polar hydrogen atoms and Kollman charges. The 3D structures of the ligands were also loaded into AutoDock Tools (Morris et al., 2009), where nonpolar hydrogen atoms were removed and torsion tree root was detected. With the ligands treated as flexible with rotatable bonds while the receptor protein treated as rigid, ligand docking of the protein at the binding site was performed using AutoDock Vina. Docking results of ligand–receptor complexes with each receptor were ranked according to the calculated binding affinities (kcal/mol).

To further evaluate binding free energy of the top-ranked ligand–receptor complex, we calculated its binding free energy using Molecular Mechanics Poisson–Boltzmann Surface Area (MM-PBSA) method (Ganoth et al., 2006; Huang et al., 2020; Kollman et al., 2000; Suever et al., 2008; Wang et al., 2001), which was implemented using the AMBER 18 MD Software Package (Case et al., 2018). MM-PBSA binding free energy was calculated by finding the energy difference between the complex and the receptor and ligand. The binding free energy (ΔG_binding) was calculated by summing the molecular mechanics energy change including the electrostatic energy (ΔE_elec) and the van der Waals energy (ΔE_vdW), and the solvation free energy including a polar part (ΔG_pol) and a nonpolar part (ΔG_nonpol).

MD simulation and binding free energy analyses of 1,25(OH)₂D3 in complex with ACE2, SARS-CoV-2 RBD and TMPRSS2

The classical active form of vitamin D3, 1,25(OH)₂D3, regulates calcium body homeostasis and has important pleiotropic effects on multiple body functions including immune system (Bikle, 2008, 2010a, 2010b, 2011; Holick, 2003; 2007; Plum & DeLuca, 2010). To determine the molecular and structural basis of 1,25(OH)₂D3 to inhibit ACE2 interaction with SARS-CoV-2 RBD and, to attenuate the activity of TMPRSS2, we performed 100 ns MD simulation for six systems: complexes of 1,25(OH)₂D3 with SARS-CoV2-RBD, ACE2 and TMPRSS2, and for SARS-CoV2-RBD, ACE2 and TMPRSS2 by themselves. All six 100 ns MD simulation systems were performed in duplicate for reproducibility. AMBER force field was used for the simulated systems. The AMBER ff14SB protein force field was used as parameters for the receptors. We have obtained the force field and charge parameter for 1,25(OH)₂D3 in our earlier study (Slominski et al., 2021).

MD simulations

We performed MD simulation for the studied systems using a standard protocol as described previously (Lee et al., 2008; Liu et al., 2008; Pan et al., 2011; Song et al., 2005; Suever et al., 2008; Sung et al., 2010; Wang et al., 2017; 2019; Yan et al., 2010, 2011; Yang & Song, 2016). Each complex was first minimized and solvated with TIP3P water molecules and 150 mM NaCl (physiological salt concentration) in a periodic box. At least 20 Å distance between the protein complex and the periodic box boundaries was ensured to reduce potential artifacts arising from periodicity. Additional ions of Na⁺ or Cl⁻ were added into the periodic box to neutralize the charge of protein–ligand complex. The solvated system was first energy-minimized with the protein complex and ions restrained but water unrestrained. The solvent was further equilibrated with the protein complex and ions restrained at a constant number–pressure–temperature (NPT) at 50 K and 1 atm for 10 ps. The simulated system was warmed up via constant number–volume–temperature (NVT) to 300 K by steps of 50 K lasting 10 ps each with SHAKE constraints and a time step of 2 fs. Production MD simulation of 100 ns at NPT of 300 K and 1 atm was performed. In the production simulations, electrostatics interaction was calculated using the particle-mesh Ewald method with interpolation of order 4. Lennard–Jones cutoffs set at 1.0 nm. SHAKE constraint was used on all hydrogen-heavy atom bonds to permit a dynamics time step of 2 fs. To determine the simulated systems’ equilibration tendencies, root mean square deviation (RMSD) of protein backbone atoms over time were analyzed.

MM/PBSA binding free energy analysis based on MD simulation trajectories

To better understand the binding characteristics of ACE2, SARS-CoV-2 RBD and TMPRSS2 with 1α,25(OH)₂D3, binding free energies were calculated for the complexes of 1,25(OH)₂D3 with SARS-CoV2-RBD, ACE2 and TMPRSS2. The binding free energy was determined with the MM-PBSA method as described in previous studies (Ganoth et al., 2006; Kollman et al., 2000; Suever et al., 2008; Wang et al., 2001), which was implemented using the AMBER 18 MD Software Package (https://ambermd.org/). The MM-PBSA method combines molecular mechanics, continuum electrostatics, solvent accessible surface area calculations to calculate the binding free energy from a series of snapshots obtained from the trajectories of the MD simulations. The binding free energy was calculated by finding the energy difference between the complex and the receptor and ligand (Equation (1)). The complex for each of the frames in the production trajectory was decomposed into receptor and ligand components. The resulting three components (complex, receptor and ligand) were then used for the binding free energy analysis. Binding free energy of the complex structure over the cumulative time was also used to verify the equilibration of the complex structure during the MD simulations

$$\text{Δ}G={\overline{G}}_{\text{complex}}-{\overline{G}}_{\text{receptor}}-{\overline{G}}_{\text{ligand}}$$ (1)

Snapshots at a time interval of 200 ps from the last 70 ns MD simulation trajectories after the system equilibration were used to calculate all components of binding free energy. The means and standard deviations of the binding free energy over the MD simulation trajectories were calculated after the system reached initial equilibration using the bootstrap statistical method (Efron & Tibshirani, 1998).

Conformation, dynamical motion and electrostatic potential analyses based on MD simulation trajectories

To determine the simulated systems’ equilibration tendencies and their convergences, RMSD of protein backbone atoms over time and binding free energy of the complex structure over the cumulative time were analyzed. Based on equilibrated MD simulation trajectories, we performed a series of analyses to better understand conformational stability, secondary structure and dynamical motion and electrostatic potential characteristic of ACE2, SARS-CoV2-RBD or TMPRSS2 in complex with 1,25(OH)₂D3.

Representative ligand–receptor complex structure

The dbscan (density-based spatial clustering of applications with noise) program (Ester et al., 1996) in Amber 18 was used for the clustering analyses of the equilibrated MD simulation trajectories. In dbscan, points are considered as part of a single cluster if there are at least some other points within a neighborhood radius ε. The dbscan algorithm also generates an average structure for the population of each cluster called a medoid. The medoid structure in the largest cluster was chosen as a representative structure for electrostatic potential analysis and to evaluate the conformation changes of TPRSS2, ACE2 and SARS-CoV-2-RBD by binding with 1,25(OH)₂D3.

System equilibration

To determine the simulated systems’ equilibration tendencies and its convergence, RMSD of protein backbone atoms over time of all systems were analyzed.

Conformational stability and secondary structure analyses

To characterize the conformational fluctuation of ACE2, SARS-CoV2-RBD and TMPRSS2 by binding with 1,25(OH)₂D3, root mean square fluctuations (RMSF) were calculated on a residue-by-residue basis, and averaged over the production simulation trajectories after the systems reached equilibrium. To characterize the change of the secondary structure of ACE2, SARS-CoV2-RBD and TMPRSS2 by binding with 1,25(OH)₂D3, we evaluated the secondary structure using the DSSP method in the cpptraj program of AMBER18 (Kabsch & Sander, 1983). For ACE2, SARS-CoV2-RBD or TMPRSS2 in complex with 1,25(OH)₂D3, the occupancy of each residue in the secondary structure was determined based on the percentage of time that the residue existed in the secondary structure over the equilibrated MD simulations.

Solvent accessible surface area analyses (SASA)

To determine the change of SASA for the binding areas for ACE2 and SARS-CoV-2-RBD by binding with 1,25(OH)₂D3, average SASA was calculated through the equilibrated MD simulation trajectories using the molsurf function in Amber 18. ACE2 binding surface for SARS-CoV-2 RBD includes residues 21–101 and 323–356 and SARS-CoV-2 RBD binding surface for ACE2 includes residues 470–504.

Dynamic motion analyses

To determine the dynamic motion changes of the binding surfaces between ACE2 and SARS-CoV-2-RBD by binding with 1,25(OH)₂D3, we performed principal component analysis (PCA) (Hayward et al., 1995) based on the equilibrated MD simulation trajectory. PCA was performed to decompose the motions of binding surfaces of ACE2 and SARS-CoV-2 RBD by binding with 1,25(OH)₂D3 into a few principal motions that are described by eigenvectors and eigenvalues. Eigenvalue characterizes a principal mode of the total motion of the system. The first PCA mode is usually interpreted as the direction of the largest conformational fluctuation of the system during MD simulations (Berendsen & Hayward, 2000). Porcupine plots were used to visualize the collective dynamic modes calculated from PCA analysis. To analyze the degree of correlated motions between residues for the TMPRSS2 by binding with 1,25(OH)₂D3 over the equilibrated MD simulation trajectories, we generated a dynamic cross-correlation maps (ΔCCM) to compare dynamic motion of TMPRSS2 with and without binding with 1,25(OH)₂D3.

Electrostatic potential analyses

We performed electrostatic potential analyses using APBS software (Baker et al., 2001; Jurrus et al., 2018) to evaluate the electrostatic interactions of 1,25(OH)₂D3 with ACE2, SARS-CoV-2 RBD and TMPRSS2. PDB2PQR software (Dolinsky et al., 2004, 2007) was used to convert representative complex structures of 1,25(OH)₂D3 with ACE2, SARS-CoV-2 RBD and TMPRSS2 in PDB format into PQR format that included AMBER radii and charges for each atom, which were then used to generate surface electrostatic potentials for the representative complex structure of ACE2, SARS-coV-2 RBD and TMPRSS2 with 1,25(OH)₂D3 using APBS software.

Statistical methods

To compare the differences of conformations and dynamical motions changes of ACE2, SARS-coV-2 RBD and TMPRSS2 with and without binding 1,25(OH)₂D3, both the average values and the standard deviations of the analyzed variables were calculated. Because the adjacent snapshots from the MD trajectory have the tendency to be correlated with each other, the autocorrelation times (Lee et al., 2008) for the studied variables were obtained to resample the trajectories into statistically independent periods to calculate the standard deviations for the studied variables as our previous studies (Lee et al., 2008; Liu et al., 2008; Pan and Song, 2010; Song et al., 2005; Suever et al., 2008; Wang et al., 2014; Yan et al., 2010, 2011). With the obtained decorrelation time for the studied variable, bootstrap analyses (Efron & Tibshirani, 1998) were performed to resample the trajectories to obtain statistically independent samples. The newly generated sample datasets were used to calculate the averages and the standard deviation for the interested variables. The Student’s t test (Bailey, 1995) with 95% confidence was used to determine whether the mean and standard deviations for the interested variables were significantly different. The protocol of the statistical methods used in this study has also been used in our published studies (Lee et al., 2008; Liu et al., 2008; Pan and Song, 2010; Song et al., 2005; Suever et al., 2008; Wang et al., 2014; Yan et al., 2010, 2011).

Results and discussion

Constructed structure of TMPRSS2

To construct TMPRSS2 structure, a template structure needs to be identified. Sequence alignment of TMPRSS2 with its homologous protein TMPRSS1 using T-Coffee (Chang et al., 2014; 2015; Di Tommaso et al., 2011; Notredame et al., 2000) showed that the two proteins had 31.7% sequence identity and 51.3% sequence similarity (Figure S1), making TMPRSS1 structure a good template for constructing the structure of TMPRSS2. With the crystal structure of TMPRSS1 (PDBID: 1Z8G (Herter et al., 2005), resolution: 1.55 Å) as template, the TMPRSS2 structure was constructed with MODELLER (Eswar et al., 2006). To generate the TMPRSS2 structures, an alignment was first created between TMPRSS1 and TMPRSS2. The alignment along with the TMPRSS1 PDB file were used to generate predictive structures of TMPRSS2. The quality of the constructed TMPRSS2 structures were evaluated with ERRAT (Colovos & Yeates, 1993), PROCHECK (Laskowski et al., 1993, 1996), and WHATCHECK (Hooft et al., 1996) (Table S1). The constructed TMPRSS2 structure with the best structure quality evaluation is shown as Figure 2(A) along with this structure’s Ramachandran plot (Figure 2(B)). The catalytic triad from the constructed TMPRSS2 structure, consisting of His²⁹⁶, Asp³⁴⁵ and Ser⁴⁴¹, were grouped close together with the R groups of these three residues all pointing to a central position, and the substrate binding residue Asp⁴³⁵ remains near the catalytic triad on a random loop (Figure 2(A)). Position of the catalytic triad and Asp⁴³⁵ further validates the constructed TMPRSS2 structure. The Ramachandran plot, generated by PROCHECK (Laskowski et al., 1993, 1996), shows that the majority of residues in the TMPRSS2 structure lie within the most favored region (red), with most of the other residues lying in the allowed region (Δark yellow) (Figure 2(B)). A combined 96.6% of TMPRSS2’s residues lie within the most favored and allowed regions, showing the constructed TMPRSS2 structure with good quality. Inhibiting the catalytic triad could help prevent SARS-CoV-2 from infecting a cell. This constructed structure of TMPRSS2 was used for molecular docking targeting the catalytic triad with vitamin D3 and its classical (25(OH)D3, 1,25(OH)₂D3) and CYP11α1-derived (20S(OH)D3, 1α,20S(OH)₂D3, 20S,23S(OH)₂D3, 1α,20S,23S(OH)₂D3) hydroxyderivatives and chemically synthesized 20S,23R(OH)₂D3, and for MD simulations of TMPRSS2 with 1,25(OH)₂D3.

Figure 2.

Constructed structure of TMPRSS2. (A) The active site of the catalytic triad consists of His²⁹⁶ (orange), Asp³⁴⁵ (yellow), and Ser⁴⁴¹ (blue). Substrate binding residue Asp⁴³⁵ in white. The figure was generated with PyMOL. (B) The Ramachandran plot, generated by PROCHECK (Laskowski et al., 1996; Laskowski et al., 1993), shows a vast majority of residues lying within the most favored (red) and allowed (Δark yellow) regions. Six residues lie within the generously allowed region (pale yellow) and four lie within the disallowed region (white).

Favorable binding of vitamin D3 and its hydroxyderivatives with ACE2, SARS-CoV2-RBD and TMPRSS2

After molecular docking with AutoDock Vina (Trott & Olson, 2010), the top-scored docking complexes of vitamin D3 and its hydroxyderivatives with ACE2, SARS-CoV2-RBD and TMPRSS2 were manually inspected and the binding free energy of each top-scored complex was calculated with MM-PBSA method (Ganoth et al., 2006; Huang et al., 2020; Kollman et al., 2000; Suever et al., 2008; Wang et al., 2001). In Table 1, we show the molecular docking results of vitamin D3 and its hydroxyderivatives with ACE2, SARS-CoV2-RBD and TMPRSS2 and MM-PBSA binding free energy result for top-scored docking complexes of vitamin D3 and its hydroxyderivatives with ACE2, SARS-CoV2-RBD and TMPRSS2. Results showed that vitamin D3 and its hydroxyderivatives of 1α,25(OH)₂D3, 25(OH)D3, 20S(OH)D3, 1α,20S(OH)₂D3, 20S,23R(OH)₂D3, 20S,23S(OH)₂D3, 1α,20S,23S(OH)₃D3 are all favorable to bind to the active site of TMPRSS2, further serving as TMPRSS2 inhibitor to block SARS-CoV-2 entry.

For ACE2 binding site for SARS-CoV2-RBD, vitamin D3 and its hydroxyderivatives, 1α,25(OH)₂D3, 25(OH)D3, 20S(OH)D3, 1α,20S(OH)₂D3, 20S,23S(OH)₂D3 and 1α,20S,23S(OH)₃D3 are all favorable to interact with ACE2, the exception was 20S,23R(OH)₂D3. For SARS-CoV2-RBD binding site for ACE2, except 20S(OH)D3, vitamin D3 and its hydroxyderivatives of 1α,25(OH)₂D3, 25(OH)D3, 1α,20S(OH)₂D3, 20S,23R(OH)₂D3, 20S,23S(OH)₂D3 and 1α,20S,23S(OH)₃D3 are all favorable to binding SARS-CoV2-RBD. Vitamin D3 or its hydroxyderivatives with favorable binding to the interaction sites between ACE2 and SARS-CoV2-RBD – either on ACE2 or on SARS-CoV2-RBD (Figure 3(A, B)) – is likely to prevent SARS-CoV2-RBD binding ACE2 to constrain cellular entry of SARS-CoV-2. For TMPRSS2, vitamin D3 and its hydroxyderivatives all bind to the active site of TMPRSS2, which could prevent SARS-CoV-2 from infecting a cell (Figure 3(C)). The above findings are important because CYP11α1-derived 20S(OH)D3, 20S23S(OH)₂D3 lack the calcemic effects, while 1α,20S(OH)₂D3 has low calcemic activity, which contrasts with highly calcemic 1α,25(OH)₂D3 (Chen et al., 2014; Slominski et al., 2010, 2013; Wang et al., 2012). It must be noted that CYP11-derived vitamin D3 hydroxymetabolites are non-toxic in preclinical studies (Chen et al., 2014; Slominski et al., 2010, 2013; Wang et al., 2012), are endogenously produced in the human body (Slominski et al., 2015, 2020a) and are detectable in natural products such as honey (Kim et al., 2020). This defines them as hormonal/biological regulators and/or natural products. However, they still require FDA approval to be considered as potential therapeutics for COVID-19 (Slominski et al., 2020c). Therefore, taking into consideration this constraint, we focused on the classical 1α,25(OH)₂D3 for subsequent MD simulations.

Figure 3.

Binding mode for compounds listed in Table 1 in ligand binding mode of (A) ACE2, (B) SARS-CoV-2 RBD and (C) TMPRSS2. The light blue meshing areas shown in the figure is the binding area. All ligands listed in Table 1 are shown as green (image made with PyMOL (v2.4.0, https://pymol.org/2/)).

Binding interactions analysis between ACE2, SARS-CoV2-RBD or TMPRSS2 with vitamin D3 and its derivatives

The binding interaction profile of the vitamin D3 and its hydroxyderivatives with ACE2, SARS-CoV2-RBD and TMPRSS2 was analyzed through the Maestro v12.4 (Figures 4–6). Several potential hydrogen bonds and non-bonding interactions were observed between vitamin D3 or its hydroxyderivatives and the binding site amino residues of ACE2, SARS-CoV2-RBD or TMPRSS2. 25(OH)D3’s hydroxyl group and 1α,20S,23S(OH)₃D3’s hydroxyl group on C1α site were found to be critical to interact with Lys26 and Hid34 in ACE2 by forming one hydrogen bond, respectively (Figure 4). For complex of SARS-CoV2-RBD with vitamin D3 or its hydroxyderivatives, 20S,23R(OH)₂D3 and 20(OH)D3’s hydroxyl group all formed one hydrogen bond with Tyr449, and 20S,23S(OH)₂D3’s hydroxyl group on C20S site formed one hydrogen bond with Gln474 (Figure 5). For complex of TMPRSS2 with vitamin D3 or its hydroxyderivatives, 1α,25(OH)₂D3 docked nicely into the active site of TMPRSS2, its hydroxyl group interacted with Leu239 and Ile157 through two hydrogen bonds, 20S,23S(OH)₂D3, 20(OH)D3 and 1α,20(OH)₂D3 formed one hydrogen bond with Met320, Thr309 and Asn192, respectively (Figure 6). Beyond above, several hydrophobic interactions and electrostatic interactions were observed between vitamin D3 or its hydroxyderivatives with ACE2, SARS-CoV2-RBD or TMPRSS2.

Figure 4.

2D interactions of vitamin D3 and its hydroxyderivatives with ACE2. The figure was generated with Maestro (v12.4, https://www.schrodinger.com/products/maestro).

Figure 6.

2D interactions of vitamin D3 and its hydroxyderivatives with TMPRSS2. The figure was generated with Maestro (v12.4, https://www.schrodinger.com/products/maestro).

Figure 5.

2D interactions of vitamin D3 and its hydroxyderivatives with SARS-CoV2-RBD. The figure was generated with Maestro (v12.4, https://www.schrodinger.com/products/maestro).

Equilibration of MD simulation systems

The RMSD of the backbone of ACE2, SARS-CoV-2 RBD and TMPRSS2 over 100 ns MD simulations (Figure 7(A, B & C)) reached initial equilibration after 30 ns MD simulations for all twelve systems with or without 1,25(OH)₂D3 binding. The binding free energies for all the complexes as a function of cumulative time over the 100 ns MD simulation also showed that the complexes converged after the first 30 ns MD simulation (Figure 7(D)). The last 70 ns of equilibrated MD simulation trajectories for ACE2, SARS-CoV-2 RBD and TMPRSS2 with or without 1,25(OH)₂D3 binding were used for the analyses of the binding thermodynamics of ACE2, SARS-CoV-2 RBD and TMPRSS2 with 1,25(OH)₂D3, and for determining the structural, conformational and motion characteristics of ACE2, SARS-CoV-2 RBD and TMPRSS2 by binding with 1,25(OH)₂D3.

Figure 7.

RMSD of ACE2 (A), SARS-coV-2 (B) and TMPRSS2 (C) with or without binding 1a,25(OH)₂D3 (1,25(OH)₂D3) over 100 ns MD simulations, and binding free energy of ACE2, SARS-coV-2 and TMPRSS2 with 1,25(OH)₂D3 as a function of cumulative time over the 100 ns MD simulation (D).

Binding thermodynamics and ligand–residue interaction decomposition analysis of ACE2, SARS-CoV-2 RBD and TMPRSS2 with 1,25(OH)₂D3 based on equilibrated MD simulation trajectories

Based on the equilibrated last 70 ns MD simulations trajectories performed in duplicate, we calculated the binding free energy of ACE2, SARS-CoV-2 and TMPRSS2 with 1,25(OH)₂D3 (Table 2). Energy components of electrostatic energy (ΔE_{electrostatic}), van der Waals energy (ΔE_vdW), polar solvation energy (ΔG_{polar-solvation}) and nonpolar solvation energy (ΔG_{nonpolar-solvation}) that contributed to the binding free energy were also calculated. The means of the binding free energy and energy components were obtained by averaging over duplicate trajectories, and the standard deviations were calculated with the independent periods of the MD simulation trajectories, using the bootstrap statistics method described in the ‘Materials and Methods’ section as in our previous studies (Lee et al., 2008; Liu et al., 2008; Pan and Song, 2010; Song et al., 2005; Suever et al., 2008; Wang et al., 2014; Yan et al., 2010, 2011). Results in Table 2 show that the binding free energy of ACE2, SARS-CoV-2 and TMPRSS2 with 1,25(OH)₂D3 are −18.55 ± 4.16, −16.97 ± 1.69 and −21.04 ± 1.53 kcal/mol separately, further confirming that ACE2, SARS-CoV-2 and TMPRSS2 are all favorable binding with 1,25(OH)₂D3. Although 1,25(OH)₂D3 display favorable binding with ACE2, SARS-CoV-2 and TMPRSS2, different components of binding energy such as van der Waals energy, electrostatics energy and polar solvation energy were observed between 1,25(OH)₂D3 in complex with ACE2, SARS-CoV-2 and TMPRSS2. The different structure, conformation and atomic charge parameters for ACE2, SARS-CoV-2 and TMPRSS2 directly affect 1,25(OH)₂D3 interactions with each receptor. These different interactions could result in the varied hydrogen bond formation, van der Waals interactions and electrostatic interactions, further contributing to the observed different energy components for binding free energy between different complexes of 1,25(OH)₂D3 with ACE2, SARS-CoV-2 and TMPRSS2.

Table 2.

Binding free energy of ACE2, SARS-CoV-2 and TMPRSS2 with 1,25(OH)₂D3 (kcal/mol).

	ACE2_1,25(OH)2D3	SARS-CoV-2 RBD _1,25(OH)2D3	TMPRSS2_ 1,25(OH)2D3
ΔEvdW	−26.44 ± 0.42	−24.08 ± 1.13	−38.99 ± 1.08
ΔEelectrostatic	−38.99 ± 1.08	−11.51 ± 1.08	−12.03 ± 4.90
ΔGnonpolar-solvation	−4.35 ± 0.04	−4.20 ± 0.09	−5.47 ± 0.08
ΔGpolar-solvation	22.37 ± 3.45	22.95 ± 1.12	35.52 ± 1.83
ΔGbinding	−18.55 ± 4.16	−16.97 ± 1.69	−21.04 ± 1.53

All values in this table were expressed in term of kcal/mol. ΔE_vdW, van der Waals energy; ΔE_{electrostatic}, electrostatic energy; ΔG_{nonpolar-solvation}, nonpolar solvation energy; ΔG_{polar-solvation}, polar solvation energy; ΔG_bingding, binding free energy for the complex.

To further quantify the contribution of binding pocket residues to the molecular interaction of 1,25(OH)₂D3 with ACE2, SARS-CoV-2 RBD and TMPRSS2, the energy decomposition per residue was performed (Figure 8). The plots of energy decomposition analyses show that the active site residues, Lys26, Asn90, Thr92, Val93, Gln96, Gln388, Pro389, Leu392, Arg559, Leu560 and Ser563 energetically favor the binding stability of 1,25(OH)₂D3 to ACE2. Remarkably, Thr92 contributed the highest binding energy, ΔE_vdW (−1.89kcal/mol), ΔE_{electrostatic} (−0.81kcal/mol) and ΔG_bingding (−1.67kcal/mol), which indicated the favorable electrostatic and vdW interactions with 1,25(OH)₂D3 (Figure 8(A)). The binding interaction with SARS-CoV-2 RBD shows that the amino acid Arg403, Tyr453, Leu455, Glu484, Gly485, Asn487, Cys488, Tyr489, Phe490, Gln493 and Tyr505 contributed the most to the total ΔG_bingding of −16.97 kcal/mol (Figure 8(B)). Figure 8(C) shows the energy decomposition plot of TMPRSS2 indicates the substantial contribution of amino acids, Lys193, Asp286, Ser287, Gln289, Trp312 and Gly313 to hold 1,25(OH)₂D3 at the binding pocket.

Figure 8.

Residue decomposition plots obtained from MM-PBSA that represent the binding energy contribution of active site residues in ACE2 (A), SARS-CoV-2 RBD (B) and TMPRSS2 (C) when binding with 1,25(OH)₂D3 based on the equilibrated last 70 ns MD simulations trajectories.

Interaction of 1,25(OH)₂D3 with ACE2 and its effect on structure, conformation, and dynamical motion of ACE2 binding surface for SARS-CoV-2 RBD

The representative complex structure of 1,25(OH)₂D3 with ACE2 in the equilibrated MD trajectory is shown in Figure 9(A). The interaction analyses showed that 1,25(OH)₂D3 bound ACE2 through one hydrogen bond with Gln96 and hydrophobic interaction with Val93, Pro389, Leu392 and Leu560. The important residues confining the binding site were found to be Lys26, Gln96, Asn90, Thr92, Val93, Gln388, Pro389, Leu392, Leu560, Ser563 and Arg559 (Figure 9(B, C)). These residues are all belong to the ACE2 binding region of SARS-CoV-2 RBD, which means that the conformational changes within this region could preclude the binding to the SARS-CoV-2 RBD (Gross et al., 2020). The electrostatic potential analyses showed the strong negative electrostatic potential (in red) in the surrounding of ACE binding region due to the negatively charged residues of Glu22, Glu23 Asp30 and Glu564. The binding surface with strong negative electrostatic potential can steadily hold the 1,25(OH)₂D3 polar groups that have partial positive charges through electrostatic interaction (Figure 9(D)). Based on the post-equilibrated MD simulation trajectories, conformational and secondary structural analyses showed that 1,25(OH)₂D3 interaction with ACE2 resulted in the conformational fluctuation changes of ACE2, including the ACE2 binding region for SARS-CoV-2 RBD (res 521–571) (Figure 9(E)), but not significant structural and secondary structural changes (Table S2, Figures S2 and S3(A)). 1,25(OH)₂D3 interaction with ACE2 resulted in the significantly decreased solvent accessible surface area of ACE2 binding region for SARS-CoV-2 RBD (Figure 9(F)). Overlay of the representative structure of ACE2 with and without 1,25(OH)₂D3 binding from the equilibrated MD simulation trajectories further demonstrated the conformation changes of the ACE2 binding region for SARS-CoV-2 RBD (Figure 9(G)). PCA of ACE2 binding region for RBD of SARS-CoV-2 showed that 1,25(OH)₂D3 binding ACE2 resulted in the changed dynamical motion of the ACE2 binding region for RBD of SARS-CoV-2 compared to ACE2 alone based on the first PCA mode (Figure 9(H)) and the second and the third PCA modes (Figure S4) that contributed most to the ACE2 dynamical motion. Dynamical motion change of ACE2 by binding 1,25(OH)₂D3 supported the observed changes of the conformation and solvent accessible surface area of ACE2 binding region for RBD of SARS-CoV-2.

Figure 9.

1a,25(OH)₂D3 (1,25(OH)₂D3) interactions with ACE2 resulted in the conformational and dynamical motion changes of ACE2 binding surface for SARS-CoV-2 RBD. (A) The representative ACE2 binding mode with 1,25(OH)₂D3 from equilibrated MD simulation (image made with PyMOL (v2.4.0, https://pymol.org/2/)). (B) 3D representation of 1,25(OH)₂D3 interaction with binding region of ACE2 (image made with PyMOL (v2.4.0, https://pymol.org/2/)). (C) 2D representation of 1,25(OH)₂D3 interaction with binding region of ACE2 (image made with Maestro (v12.4, https://www.schrodinger.com/products/maestro)). (D) Electrostatic potential of ACE2 at its binding region with 1,25(OH)₂D3 (in magenta). Red: negative electrostatic potential; Blue: positive electrostatic potential and White: neutral electrostatic potential (image made with PyMOL (v2.4.0, https://pymol.org/2/)). (E) RMSF of ACE2 with and without 1,25(OH)₂D3 binding (Δuplicate and data shown in average). (F) Comparison of SASA of ACE2 binding region (residues 21–101 and 323–356) for SARS-CoV-2 RBD with and without dynamic interactions with 1,25(OH)₂D3. (G) Overlay of ACE2 with and without dynamic interactions with 1,25(OH)₂D3 (Before dynamic interactions with 1,25(OH)₂D3: ACE2 shown in gray, the surface of ACE2 binding region for SARS-CoV-2 RBD shown in gray. After dynamic interactions with 1,25(OH)₂D3: ACE2 shown in orange, the surface of ACE2 binding region with SARS-CoV-2 RBD shown in red). Residues 106–130 were used for two structures alignment. (H) PCA motions of ACE2 binding region for RBD of SARS-CoV-2 with and without binding 1,25(OH)₂D3. ACE2 were shown as new cartoon in gray and its binding site for RBD of SARS-CoV-2 in red. 1,25(OH)₂D3 shown as licorice in green. Arrows of principal dynamic motion shown as blue. The images were made with VMD (v1.9.2, http://www.ks.uiuc.edu/Research/vmd/).

Interaction of 1,25(OH)₂D3 with SARS-CoV-2 RBD and its effect on conformation and dynamical motion of SARS-CoV-2 RBD binding surface for ACE2

The representative complex structure of 1,25(OH)₂D3 with SARS-CoV-2 RBD in the equilibrated MD trajectory is shown in Figure 10(A). Interaction analyses showed that 1,25(OH)₂D3 bound SARS-CoV-2 RBD through one hydrogen bond with Phe490 at C3 and two hydrogen bonds with Glu484 and Tyr489 at C1α site. 1,25(OH)₂D3 bound SARS-CoV-2 RBD also through hydrophobic interaction with Tyr453, Leu455, Cys486, Tyr489 and Phe490. The important amino residues confining the binding site were found to be Arg403, Tyr453, Leu455, Glu484, Gly485, Asn487, Cys488, Tyr489, Phe490, Gln493 and Tyr505 (Figure 10(B, C)). Electrostatic potential analyses showed both positive (in blue) and negative electrostatic potential (in red) in SARS-CoV-2 RBD binding region for ACE2 (Figure 10(D)). Positive electrostatic potential due to positive charged residues of Arg403 and Lys417 can hold 1,25(OH)₂D3 polar groups that have partial negative charges. In the meanwhile, negative charged Glu406 and Glu484 can hold 1,25(OH)2D3 polar groups that have partial positive charges through electrostatic interactions (Figure 10(D)). Based on the post-equilibrated MD simulation trajectories, conformational and secondary structural analyses showed that 1,25(OH)₂D3 interaction with SARS-CoV-2 RBD resulted in the increased conformational fluctuation for SARS-CoV-2 RBD binding surface for ACE2 (res 470–504) (Figure 10(E)), but not significant structural and secondary structural changes (Table S2, Figures S2 and S3(B)). 1,25(OH)₂D3 interaction with SARS-CoV-2 RBD resulted in the significantly decreased solvent accessible surface area of SARS-CoV-2 RBD binding region for ACE2 (Figure 10(F)). Overlay of the representative structure of SARS-CoV-2 RBD with and without 1,25(OH)₂D3 binding from the equilibrated MD simulation trajectories further demonstrated the conformation changes of the SARS-CoV-2 RBD binding region for ACE2 (Figure 10(G)). Principal component analyses of the dynamical motion of SARS-CoV-2 RBD binding surface for ACE2 showed that 1,25(OH)₂D3 binding SARS-CoV-2 RBD resulted in the changed dynamical motion of the SARS-CoV-2 RBD (Figure 10(H) and Figure S5), which support the observed changes of the conformation and solvent accessible surface area of SARS-CoV-2 RBD binding region for ACE2.

Figure 10.

1a,25(OH)₂D3 (1,25(OH)₂D3) interactions with SARS-CoV-2 RBD resulted in the conformation and dynamical motion changes of SARS-CoV-2 RBD binding surface for ACE2. (A) The representative SARS-CoV-2 RBD binding mode with 1,25(OH)₂D3 from equilibrated MD simulation (image made with PyMOL (v2.4.0, https://pymol.org/2/)). (B) 3D representation of 1,25(OH)₂D3 interaction with binding region of SARS-CoV-2 RBD (image made with PyMOL (v2.4.0, https://pymol.org/2/)). (C) 2D representation of 1,25(OH)₂D3 interaction with binding region of SARS-CoV-2 RBD (image made with Maestro (v12.4, https://www.schrodinger.com/products/maestro)). (D) Electrostatic potential of SARS-CoV-2 RBD at its binding region with 1,25(OH)₂D3 (in magenta). Red: negative electrostatic potential; Blue: positive electrostatic potential and white: neutral electrostatic potential (image made with PyMOL (v2.4.0, https://pymol.org/2/)). (E) RMSF of SARS-CoV-2 RBD with and without dynamic interactions with 1,25(OH)₂D3 binding (Δuplicate and data shown in average). (F) Comparison of SASA of SARS-CoV-2 RBD binding region (residues 470–504) for ACE2 with and without dynamic interactions with 1,25(OH)₂D3. (G) Overlay of SARS-CoV-2 RBD with and without dynamic interactions with 1,25(OH)₂D3 (without 1,25(OH)₂D3 binding: SARS-CoV-2 RBD shown in gray, the surface of ACE2 binding region with RBD of SARS-CoV-2 shown in gray. without 1,25(OH)₂D3 binding, SARS-CoV-2 RBD shown in cyan, the surface of ACE2 binding region with RBD of SARS-CoV-2 shown in blue) with the alignment using residues 506–518. (H) PCA motions of SARS-CoV-2 RBD binding region for ACE2 with and without dynamic interactions with 1,25(OH)₂D3. SARS-CoV-2 RBD were shown as new cartoon in gray and its binding with RBD of SARS-CoV-2 in blue. 1,25(OH)₂D3 shown as licorice in green. Arrows of principal dynamic motion shown as red. The images were made with VMD (v1.9.2, http://www.ks.uiuc.edu/Research/vmd/).

Interaction of 1,25(OH)₂D3 with TMPRSS2 and its effect on conformation and dynamical motion of TMPRSS2

The representative complex structure of 1,25(OH)₂D3 with TMPRSS2 in the equilibrated MD trajectory is shown as Figure 11(A). Interaction analyses showed that 1,25(OH)₂D3 bound TMPRSS2 through hydrophobic interactions with Tyr337, Trp461, Val473 and Tyr474. The important amino residues confining the binding site were found to be Hie296, Tyr337, Ser339, Lys342, Asp435, Ser436, Gly439, Thr459, Ser460, Trp461, Gly462, Ser463, Gly464, Gly472, Gln438, Val473 and Tyr474 (Figure 11(B, C)). Electrostatic potential analyses showed the strong negative electrostatic potential (in red) in the binding region of TMPRSS2 due to the negatively charged residues of Glu299, Asp338, Asp345, Glu389, Asp435, Asp440 and Asp458. The binding surface with strong negative electrostatic potential can steadily hold the 1,25(OH)₂D3 polar groups that have partial positive charges through electrostatic interaction (Figure 11(D)). Based on the post-equilibrated MD simulation trajectories, conformational and secondary structural analyses showed that 1,25(OH)₂D3 interaction with TMPRSS2 resulted in the increased conformational fluctuation, particularly for residues between 210 and 240 that are near the N-terminal of TMPRSS2 (Figure 11(E)), but not significant structural and secondary structural change (Table S2, Figures S2 and S3(C)). To determine whether TMPRSS2 conformational changes is associated with dynamical motions changes of TMPRSS2, dynamic cross-correlation maps (ΔCCM) between all Ca atoms from the backbone of TMPRSS2 with and without dynamic interactions with 1,25(OH)₂D3 were analyzed. DCCM results showed that 1,25(OH)₂D3 interaction with TMPRSS2 caused overall increased both correlated motion (red) and anti-correlated motion (blue) compared to TMPRSS2 alone (Figure 11(F)).

Figure 11.

1a,25(OH)₂D3 (1,25(OH)₂D3) interactions with TMPRSS2 resulted in the conformational and dynamical motion changes. (A) The representative TMPRSS2 binding mode with 1,25(OH)₂D3 from equilibrated MD simulation (image made with PyMOL (v2.4.0, https://pymol.org/2/)). (B) 3D representation of 1,25(OH)₂D3 interaction with binding region of TMPRSS2 (image made with PyMOL (v2.4.0, https://pymol.org/2/)). (C) 2D representation of 1,25(OH)₂D3 interaction with binding region of TMPRSS2 (image made with Maestro (v12.4, https://www.schrodinger.com/products/maestro)). (D) Electrostatic potential of TMPRSS2 at its binding region with 1,25(OH)₂D3 (in magenta). Red: negative electrostatic potential; Blue: positive electrostatic potential and White: neutral electrostatic potential. (image made with PyMOL (v2.4.0,https://pymol.org/2/)). (E) RMSF of TMPRSS2 with and without 1,25(OH)₂D3 binding (Duplicate and data shown in average). (F) DCCM dynamical motion analyses of TMPRSS2 with and without binding with 1,25(OH)₂D3 (correlated motion shown in read, anti-correlated motion shown in blue). The images were made with MATLAB (vR2015a, https://www.mathworks.com/products/matlab.html).

Conclusion

The epidemiologic correlations of SARS-CoV-2 infection and poor prognosis of disease with vitamin D deficiency are observed worldwide, however, their molecular mechanisms are not fully understood. In this study, we used combined molecular docking, MD simulations and binding free energy analyses to investigate the potentials of vitamin D3 and its hydroxyderivatives as TMPRSS2 inhibitor and to inhibit the SARS-CoV-2 RBD binding to ACE2, as well as to unveil molecular and structural basis of 1,25(OH)₂D3 capability to inhibit ACE2 and SARS-CoV-2 RBD interactions. The results have shown that vitamin D3 and its hydroxyderivatives are favorable to bind active site of TMPRSS2 and the binding site(s) between ACE2 and SARS-CoV2-RBD, which indicate that vitamin D3 and its biologically active hydroxyderivatives can serve as TMPRSS2 inhibitors and can inhibit ACE2 binding of SARS-CoV-2 RBD to prevent SARS-CoV-2 entry. The interactions of classical active form of vitamin D3, 1,25(OH)₂D3, with SARS-CoV-2 RBD and ACE2 resulted in the conformation and dynamical motion changes of the binding surfaces between SARS-CoV-2 RBD and ACE2 to interrupt the binding of SARS-CoV-2 RBD with ACE2. The interaction of 1,25(OH)₂D3 with TMPRSS2 also caused the conformational and dynamical motion changes of TMPRSS2, which could affect TMPRSS2 to prime SARS-CoV-2 spike proteins. Therefore, we propose that vitamin D3 and its biologically active hydroxyderivatives are promising drugs or adjuvants in the treatment of COVID-19 or to use their backbones for further modification to develop efficient drugs through medicinal chemistry approach.

Abbreviations:

ACE2

angiotensin-converting enzyme 2

COVID-19

coronavirus disease 2019

RBD

receptor binding domain

SARS-CoV-2

severe acute respiratory syndrome coronavirus

SASA

solvent accessible surface area

TMPRSS2

transmembrane serine protease 2

MD

molecular dynamics

MM-PBSA

molecular mechanics Poisson–Boltzmann surface area

NPT

constant number–pressure–temperature

NVT

constant number–volume–temperature

RMSD

root mean square deviation

RMSF

root mean square fluctuations

SASA

solvent accessible surface area analyses

DCCM

dynamic cross-correlation map

PCA

principal component analysis

dbscan

density-based spatial clustering of applications with noise