Generation of predictive pharmacophore model for SARS-coronavirus main proteinase.

Pharmacophore-based virtual screening is an effective, inexpensive and fast approach to discovering useful starting points for drug discovery. In this study, we developed a pharmacophore model for the main proteinase of severe acute respiratory syndrome coronavirus (SARS-CoV). Then we used this pharmacophore model to search NCI 3D database including 250, 251 compounds and identified 30 existing drugs containing the pharmacophore query. Among them are six compounds that already exhibited anti-SARS-CoV activity experimentally. This means that our pharmacophore model can lead to the discovery of potent anti-SARS-CoV inhibitors or promising lead compounds for further SARS-CoV main proteinase inhibitor development.

Introduction

The infection of the newly emerged severe acute respiratory syndrome coronavirus (SARS-CoV) is characterized by acute flu-like symptoms that progress to acute lung injury or acute respiratory distress syndrome with over 10% of mortality [1]. To date there are no universally recommended therapy for the disease. Many scientists are now making efforts to develop effective drugs against SARS. The combination therapy of corticosteroid with lopinavir, ribavirin and ritonavir can improve clinical response and reduce mortality rates apparently [2], [3]. Cinatl et al. [4] found that ribavirin, azauridine, pyrazofurin and glycyrrhizin are active against SARS-CoV. Barnard et al. [5] reported that calpain inhibitors and β-d-N4-hydroxycytidine exhibit inhibitory effects on SARS-CoV.

Structure-based drug design focuses on two important approaches: one is receptor-based docking technique, another is pharmacophore-based virtual screening technique. A pharmacophore is the 3D arrangement of atoms or functional groups essential for the compound to bind to a specific receptor [6]. The power of a pharmacophore model is to discover new leads by using 3D database pharmacophore searching and guide chemists to synthesize new compounds [7]. Such a pharmacophore-based method has been successfully applied to many drug development programs [8], [9], [10], [11], [12], [13], [14], [15], [16], [17], [18]. The main proteinase of SARS-CoV plays an important role in virus replication and is the primary target for drugs. The aim of this study is to develop 3D pharmacophore models for SARS-CoV main proteinase and expect to provide useful knowledge for anti-SARS drug design.

Material and methods

There are two methods to derive a reasonable pharmacophore model. One is from the crystal structure of protein–ligand complex, another is based on molecular modeling of enzyme with its potential inhibitors. Here we used the experimental structure of SARS-CoV main proteinase complexed with its peptide inhibitor CMK (PDB ID 1UK4) [19] and the predicted structures of SARS-CoV main proteinase with six drugs/compounds [20] for establishing pharmacophore models. The structures of CMK peptide and six compounds are shown in Fig. 1 .

Fig. 1

The peptide and compounds used for pharmacophore generation.

The POCKET module in LigBuilder program [21] was employed to obtain the pharmacophore models of SARS-CoV main proteinase. This approach was successfully applied to the identification of novel inhibitors for alanine racemase [22]. The proposed pharmacophore model is a binding-site-derived pharmacophore model, which includes the following pharmacophore features of ligands binding to the enzyme’s active site: a positively charged nitrogen atom (ammonium cation) to represent a hydrogen bond donor (HBD), a negatively charged oxygen atom (as in a carboxyl group) to represent a hydrogen bond acceptor (HBA), and a carbon atom (methane) to represent a hydrophobic center (HPC). A pharmacophore model is generated for each protein–ligand complex.

Results and discussion

Using CMK peptide and six compounds in Fig. 1, we generated a set of seven eight-point pharmacophore models for SARS-CoV main proteinase, which is listed in Table 1 . These hypotheses exhibit different features due to the diversities of the compounds involved. Based on these models, we extracted a common four-point pharmacophore distance pattern shown in Fig. 2 , P1 is HBA, HBD and HPC, P2 is HBA and HPC, P3 is HBA and HBD, P4 is HBA and HBD. Such a pharmacophore distance pattern was subsequently used for the 3D database search.

Table 1

The eight-point pharmacophore models obtained by LigBuilder from seven peptide and compounds

Pharmacophore	Peptide/compounds	Features
1	CMK peptide	HBD HBD HPC HBD HPC HBD HPC HBA
2	Lopinavir	HBA HBD HBA HBD HBA HBA HBA HBA
3	Ritonavir	HPC HPC HBD HBD HBD HBA HBD HBA
4	Niclosamide	HBA HBD HBA HBA HBA HBD HPC HBA
5	Promazine	HBA HBD HBA HPC HBA HBA HBA HBA
6	PNU	HBA HPC HBA HBD HBD HBA HBA HBA
7	UC2	HPC HBD HBD HBA HBD HPC HPC HBA

HBA = hydrogen bond acceptor, HBD = hydrogen bond donor, HPC = hydrophobic center.

Fig. 2

Four-point pharmacophore distance pattern for SARS-CoV main proteinase. Here, P1 is HBA, HBD and HPC, P2 is HBA and HPC, P3 is HBA and HBD, P4 is HBA and HBD.

The pharmacophore searching in 3D database was conducted in the 3D NCI database, which has 250, 251 open structures ready for searching. Table 2 summarizes the results for similarity search of above-mentioned four-point pharmacophore model with constraints: (1) the distance ranges for P1P2, P1P3, P2P3, P1P4, P2P4 and P3P4 (Fig. 2) are 2–3, 2–3, 1.5–2.5, 5–6, 3–4 and 3–4 Å, respectively; (2) the compound is drug; (3) the antiviral probability is over 70%; (4) the compounds labeled as “No Name” is not included. After review of these hitlists, 30 drugs were selected for further analysis, their chemical structures and bioactivities (documented in anti-HIV and anti-opportunistic infection chemical compound database, which contains approximately 100,000 compounds, http://www.apps1.niaid.nih.gov/struct_search/an/an_search.htm) are shown in Fig. 3 and Table 3 , respectively. It is noted that almost all drugs exhibited anti-HIV activity, and some of them have activities against Mycobacterium tuberculosis (puromycin), influenza virus (5-bromo-2′-deoxycytidine), hepatitis virus (dideoxyguanosine, 2′,3′-dideoxycytidine and ribavirin), dengue virus (azauridine and ribavirin), rhinovirus and poliovirus (ribavirin). This should be very meaningful in consideration of the following facts: (1) there are some links between SARS-CoV and HIV and HBV [23], [24], [25]; (2) similar structure patterns exist in SARS-CoV main proteinase with rhinovirus 3c protease, poliovirus 3c proteinase, HAV 3c protease, HCV Ns3 protease and dengue virus Ns3 protease [26]; (3) SARS-CoV has clinically similar symptoms with influenza virus/M. tuberculosis, such as fever, cough, pains, pneumonia and death [27].

Table 2

Summary of NCI database search by four-point pharmacophores

Pharmacophore features in Fig. 2				Hits
P1	P2	P3	P4
HBD	HBA	HBD	HBD	987
HBA	HBA	HBA	HBA	794
HPC	HBA	HBA	HBD	286
HPC	HPC	HBA	HBD	305
HPC	HBA	HBD	HBD	298
HBA	HBA	HBD	HBD	881
HBA	HBA	HBA	HBD	895

HBA = hydrogen bond acceptor, HBD = hydrogen bond donor, HPC = hydrophobic center.

Fig. 3

Chemical structures of 30 drugs obtained by four-point pharmacophore search in NCI 3D database.

Table 3

Thirty drugs obtained by four-point pharmacophore search in NCI 3D database

Name	NSC number	Formula	Bioactivity documented in HIV/OI therapeutics database (http://www.apps1.niaid.nih.gov/struct_search/an/an_search.htm)
PD-ADI	218321	C11H16N4O4	HIV
Coformycin	277817	C11H16N4O5	HIV
Zidovudine	602670	C10H13N5O4	HIV, HSV, human cytomegalovirus, vaccinia virus, cowpox virus
Vira-A	404241	C10H13N5O4	HIV, HSV, human cytomegalovirus, varicella-zoster virus, vaccinia virus, cowpox virus
Angustmycin C	53104	C11H15N5O5
ARA-AMP	259272	C10H14N5O7P	HIV, HSV, vaccinia virus
Cordycepin	63984	C10H13N5O3	HIV
Triciribine	154020	C13H16N6O4	HIV, human cytomegalovirus, HSV
5-AZCR	102816	C8H12N4O5	HIV
Puromycin	3055	C22H29N7O5	HIV, M. tuberculosis
CHETOMIN	289491	C31H30N6O6S4	HIV
Vengicide	99843	C12H13N5O4	Human cytomegalovirus
Spongothymidin	68929	C10H14N2O6	HIV, HSV, varicella-zoster virus
Arauridine	68928	C9H12N2O6
P-Ara-C	135962	C25H43N3O6	HIV
Pyrazofurin	143095	C9H13N3O6	HIV, vaccinia virus, West Nile virus
Thymidin	21548	C10H14N2O5	HIV, varicella-zoster virus
Radibud	38297	C9H11BrN2O5
Alexan	63878	C9H13N3O5	HIV, HSV, human cytomegalovirus, varicella-zoster virus, measles virus
Floxuridin	27640	C9H11FN2O5	HIV, HSV, vaccinia virus
Gemcitabine	613327	C9H12ClF2N3O4	HIV, cowpox virus, vaccinia virus
Dideoxyguanosine	619072	C10H13N5O3	HIV, HBV
2′,3′-Dideoxycytidine	606170	C9H13N3O3	HIV, HBV
5-Bromo-2′-deoxycytidine	61765	C9H12BrN3O4	HSV
Ribavirin	163039	C8H12N4O5	HIV, HSV, HCV, influenza virus, dengue virus, measles virus, respiratory syncytial virus, rhinovirus, polio virus, vaccinia virus, cowpox virus
Azauridine	32074	C8H11N3O6	West Nile virus, cowpox virus, vaccinia virus, dengue virus, Japanese encephalitis virus, Yellow fever virus
Fialuridine	678514	C9H10FIN2O5	Cowpox virus, vaccinia virus, HSV, varicella-zoster virus
Emanil	39661	C9H11IN2O5	HIV, HSV, cowpox virus, vaccinia virus, varicella-zoster virus
Tubercidin	56408	C11H14N4O4	HIV, vaccinia virus, human cytomegalovirus
Viroptic	75520	C10H11F3N2O5	HSV, cowpox virus, vaccinia virus

Indeed, among the 30 drugs are six compounds that already exhibited anti-SARS-CoV activity experimentally: azauridine, pyrazofurin, ribavirin, 2′,3′-dideoxycytidine, dideoxyguanosine, and 5-bromo-2′-deoxycytidine [4], [5]. This shows that our pharmacophore model can lead to the discovery of potent anti-SARS-CoV inhibitors or at least provide some useful clues. Fig. 4 shows the mappings of the six compounds into the four-point pharmacophore model, which are mapped to 1–3 HBA, 1–2 HBD and 0–1 HPC. In addition, most of the remaining compounds have remarkable similarities with one of the above six compounds, for example, azauridine with 5-AZCR, alexan, arauridine and spongothymidin; 5-bromo-2′-deoxycytidine with emanil, fialuridine, floxuridin, radibud, thymidin and viroptic; dideoxyguanosine with angustmycin C, cordycepin, tubercidin, vengicide and vira-A. The superpositions for these compounds are shown in Fig. 5 . In summary, our results indicate that the existing 30 drugs identified by our pharmacophore model could be potential inhibitors against SARS-CoV, or at least good lead compounds for anti-SARS-CoV drug design.

Fig. 4

The mappings of six compounds that experimentally exhibited anti-SARS-CoV activity into the four-point pharmacophore model: (A) azauridine, (B) 5-bromo-2′-deoxycytidine, (C) dideoxycytidine, (D) dideoxyguanosine, (E) pyrazofurin, (F) ribavirin.

Fig. 5

Superpositions of anti-SARS-CoV compounds with other compounds. (A) azauridine with 5-AZCR, alexan, arauridine and spongothymidin; (B) 5-bromo-2′-deoxycytidine with email, fialuridine, floxuridin, radibud, thymidin and viroptic; (C) dideoxyguanosine with angustmycin C, cordycepin, tubercidin, vengicide and vira-A.